Datasets:

---

bigcode

/

jupyter-parsed

like 4

Follow

BigCode 1.35k

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "### Import Libraries", "_____no_output_____" ] ], [ [ "import sys\n!{sys.executable} -m pip install -r requirements.txt", "Requirement already satisfied: mysqlclient in /Users/shawlu/anaconda3/lib/python3.5/site-packages (from -r requirements.txt (line 1)) (1.3.13)\r\n" ], [ "import numpy as np\nimport matplotlib.pyplot as plt\nfrom analytics import SQLClient", "_____no_output_____" ] ], [ [ "### Connect to MySQL database", "_____no_output_____" ] ], [ [ "username = \"privateuser\"\npassword = \"1234567\"\nport = 7777\n\nclient = SQLClient(username, password, port)", "_____no_output_____" ], [ "sql_tmp = \"\"\"\n SELECT \n id\n ,userID\n ,name\n ,type\n ,-priceCNY * count / 6.9 AS price\n ,count\n ,currency\n ,-priceCNY * count AS priceCNY\n ,time \n FROM items \n WHERE userID LIKE '%%shawlu%%' \n AND time BETWEEN '$start_dt$ 00:00:00' AND '$end_dt$ 00:00:00' \n AND deleted = 0\n ORDER BY time;\"\"\"", "_____no_output_____" ] ], [ [ "### Analytics: Nov. 2018", "_____no_output_____" ] ], [ [ "start_dt = '2018-11-01'\nend_dt = '2018-12-01'\ndf = client.query(sql_tmp.replace('$start_dt$', start_dt).replace(\"$end_dt$\", end_dt))", "_____no_output_____" ], [ "df = df.groupby(['type']).sum()\ntotal = np.sum(df.price)\ndf[\"pct\"] = df.price / total\ndf[\"category\"] = client.categories\ndf = df.sort_values(\"pct\")[::-1]\ndf", "_____no_output_____" ], [ "labels = [\"%s: $%.2f\"%(df.category.values[i],\n df.price.values[i]) for i in range(len(df))]\n\nplt.figure(figsize=(8, 8))\n\ntitle = \"Total expense %s\\n%s-%s\"%('$ {:,.2f}'.format(total), start_dt, end_dt)\nplt.title(title)\n\n_ = plt.pie(x = df.price.values, \n labels = labels,\n autopct='%1.1f%%',\n labeldistance = 1.1)\n\ncentre_circle = plt.Circle((0,0),0.70,fc='white')\nfig = plt.gcf()\nfig.gca().add_artist(centre_circle)\n\nplt.savefig(\"month.png\")", "_____no_output_____" ] ], [ [ "### Analytics: Year of 2018", "_____no_output_____" ] ], [ [ "start_dt = '2018-01-01'\nend_dt = '2018-12-01'\ndf = client.query(sql_tmp.replace('$start_dt$', start_dt).replace(\"$end_dt$\", end_dt))", "_____no_output_____" ], [ "df[df.type == 'COM']", "_____no_output_____" ], [ "df = df.groupby(['type']).sum()\ntotal = np.sum(df.price)\ndf[\"pct\"] = df.price / total\ndf[\"category\"] = client.categories\ndf = df.sort_values(\"pct\")[::-1]\ndf", "_____no_output_____" ], [ "labels = [\"%s: $%.2f\"%(df.category.values[i],\n df.price.values[i]) for i in range(len(df))]\n\nplt.figure(figsize=(8, 8))\n\ntitle = \"Total expense %s\\n%s-%s\"%('$ {:,.2f}'.format(total), start_dt, end_dt)\nplt.title(title)\n\n_ = plt.pie(x = df.price.values, \n labels = labels,\n autopct='%1.1f%%',\n labeldistance = 1.1)\n\ncentre_circle = plt.Circle((0,0),0.70,fc='white')\nfig = plt.gcf()\nfig.gca().add_artist(centre_circle)\n\nplt.savefig(\"year.png\")", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "## Dependencies", "_____no_output_____" ] ], [ [ "import os\nimport sys\nimport cv2\nimport shutil\nimport random\nimport warnings\nimport numpy as np\nimport pandas as pd\nimport seaborn as sns\nimport multiprocessing as mp\nimport matplotlib.pyplot as plt\nfrom tensorflow import set_random_seed\nfrom sklearn.utils import class_weight\nfrom sklearn.model_selection import train_test_split\nfrom sklearn.metrics import confusion_matrix, cohen_kappa_score\nfrom keras import backend as K\nfrom keras.models import Model\nfrom keras.utils import to_categorical\nfrom keras import optimizers, applications\nfrom keras.preprocessing.image import ImageDataGenerator\nfrom keras.layers import Dense, Dropout, GlobalAveragePooling2D, Input\nfrom keras.callbacks import EarlyStopping, ReduceLROnPlateau, Callback, LearningRateScheduler, ModelCheckpoint\n\ndef seed_everything(seed=0):\n random.seed(seed)\n os.environ['PYTHONHASHSEED'] = str(seed)\n np.random.seed(seed)\n set_random_seed(0)\n\nseed = 0\nseed_everything(seed)\n%matplotlib inline\nsns.set(style=\"whitegrid\")\nwarnings.filterwarnings(\"ignore\")\nsys.path.append(os.path.abspath('../input/efficientnet/efficientnet-master/efficientnet-master/'))\nfrom efficientnet import *", "/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:516: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n _np_qint8 = np.dtype([(\"qint8\", np.int8, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:517: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n _np_quint8 = np.dtype([(\"quint8\", np.uint8, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:518: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n _np_qint16 = np.dtype([(\"qint16\", np.int16, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:519: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n _np_quint16 = np.dtype([(\"quint16\", np.uint16, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:520: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n _np_qint32 = np.dtype([(\"qint32\", np.int32, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:525: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n np_resource = np.dtype([(\"resource\", np.ubyte, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:541: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n _np_qint8 = np.dtype([(\"qint8\", np.int8, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:542: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n _np_quint8 = np.dtype([(\"quint8\", np.uint8, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:543: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n _np_qint16 = np.dtype([(\"qint16\", np.int16, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:544: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n _np_quint16 = np.dtype([(\"quint16\", np.uint16, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:545: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n _np_qint32 = np.dtype([(\"qint32\", np.int32, 1)])\n/opt/conda/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:550: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n np_resource = np.dtype([(\"resource\", np.ubyte, 1)])\nUsing TensorFlow backend.\n" ] ], [ [ "## Load data", "_____no_output_____" ] ], [ [ "fold_set = pd.read_csv('../input/aptos-split-oldnew/5-fold.csv')\nX_train = fold_set[fold_set['fold_2'] == 'train']\nX_val = fold_set[fold_set['fold_2'] == 'validation']\ntest = pd.read_csv('../input/aptos2019-blindness-detection/test.csv')\n\n# Preprocecss data\ntest[\"id_code\"] = test[\"id_code\"].apply(lambda x: x + \".png\")\n\nprint('Number of train samples: ', X_train.shape[0])\nprint('Number of validation samples: ', X_val.shape[0])\nprint('Number of test samples: ', test.shape[0])\ndisplay(X_train.head())", "Number of train samples: 18697\nNumber of validation samples: 733\nNumber of test samples: 1928\n" ] ], [ [ "# Model parameters", "_____no_output_____" ] ], [ [ "# Model parameters\nmodel_path = '../working/effNetB4_img256_noBen_fold3.h5'\nFACTOR = 4\nBATCH_SIZE = 8 * FACTOR\nEPOCHS = 20\nWARMUP_EPOCHS = 5\nLEARNING_RATE = 1e-3/2 * FACTOR\nWARMUP_LEARNING_RATE = 1e-3/2 * FACTOR\nHEIGHT = 256\nWIDTH = 256\nCHANNELS = 3\nTTA_STEPS = 5\nES_PATIENCE = 5\nLR_WARMUP_EPOCHS = 5\nSTEP_SIZE = len(X_train) // BATCH_SIZE\nTOTAL_STEPS = EPOCHS * STEP_SIZE\nWARMUP_STEPS = LR_WARMUP_EPOCHS * STEP_SIZE", "_____no_output_____" ] ], [ [ "# Pre-procecess images", "_____no_output_____" ] ], [ [ "old_data_base_path = '../input/diabetic-retinopathy-resized/resized_train/resized_train/'\nnew_data_base_path = '../input/aptos2019-blindness-detection/train_images/'\ntest_base_path = '../input/aptos2019-blindness-detection/test_images/'\ntrain_dest_path = 'base_dir/train_images/'\nvalidation_dest_path = 'base_dir/validation_images/'\ntest_dest_path = 'base_dir/test_images/'\n\n# Making sure directories don't exist\nif os.path.exists(train_dest_path):\n shutil.rmtree(train_dest_path)\nif os.path.exists(validation_dest_path):\n shutil.rmtree(validation_dest_path)\nif os.path.exists(test_dest_path):\n shutil.rmtree(test_dest_path)\n \n# Creating train, validation and test directories\nos.makedirs(train_dest_path)\nos.makedirs(validation_dest_path)\nos.makedirs(test_dest_path)\n\ndef crop_image(img, tol=7):\n if img.ndim ==2:\n mask = img>tol\n return img[np.ix_(mask.any(1),mask.any(0))]\n elif img.ndim==3:\n gray_img = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)\n mask = gray_img>tol\n check_shape = img[:,:,0][np.ix_(mask.any(1),mask.any(0))].shape[0]\n if (check_shape == 0): # image is too dark so that we crop out everything,\n return img # return original image\n else:\n img1=img[:,:,0][np.ix_(mask.any(1),mask.any(0))]\n img2=img[:,:,1][np.ix_(mask.any(1),mask.any(0))]\n img3=img[:,:,2][np.ix_(mask.any(1),mask.any(0))]\n img = np.stack([img1,img2,img3],axis=-1)\n \n return img\n\ndef circle_crop(img):\n img = crop_image(img)\n\n height, width, depth = img.shape\n largest_side = np.max((height, width))\n img = cv2.resize(img, (largest_side, largest_side))\n\n height, width, depth = img.shape\n\n x = width//2\n y = height//2\n r = np.amin((x, y))\n\n circle_img = np.zeros((height, width), np.uint8)\n cv2.circle(circle_img, (x, y), int(r), 1, thickness=-1)\n img = cv2.bitwise_and(img, img, mask=circle_img)\n img = crop_image(img)\n\n return img\n \ndef preprocess_image(image_id, base_path, save_path, HEIGHT=HEIGHT, WIDTH=WIDTH, sigmaX=10):\n image = cv2.imread(base_path + image_id)\n image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)\n image = circle_crop(image)\n image = cv2.resize(image, (HEIGHT, WIDTH))\n# image = cv2.addWeighted(image, 4, cv2.GaussianBlur(image, (0,0), sigmaX), -4 , 128)\n cv2.imwrite(save_path + image_id, image)\n \ndef preprocess_data(df, HEIGHT=HEIGHT, WIDTH=WIDTH, sigmaX=10):\n df = df.reset_index()\n for i in range(df.shape[0]):\n item = df.iloc[i]\n image_id = item['id_code']\n item_set = item['fold_2']\n item_data = item['data']\n if item_set == 'train':\n if item_data == 'new':\n preprocess_image(image_id, new_data_base_path, train_dest_path)\n if item_data == 'old':\n preprocess_image(image_id, old_data_base_path, train_dest_path)\n if item_set == 'validation':\n if item_data == 'new':\n preprocess_image(image_id, new_data_base_path, validation_dest_path)\n if item_data == 'old':\n preprocess_image(image_id, old_data_base_path, validation_dest_path)\n \ndef preprocess_test(df, base_path=test_base_path, save_path=test_dest_path, HEIGHT=HEIGHT, WIDTH=WIDTH, sigmaX=10):\n df = df.reset_index()\n for i in range(df.shape[0]):\n image_id = df.iloc[i]['id_code']\n preprocess_image(image_id, base_path, save_path)\n\nn_cpu = mp.cpu_count()\ntrain_n_cnt = X_train.shape[0] // n_cpu\nval_n_cnt = X_val.shape[0] // n_cpu\ntest_n_cnt = test.shape[0] // n_cpu\n\n# Pre-procecss old data train set\npool = mp.Pool(n_cpu)\ndfs = [X_train.iloc[train_n_cnt*i:train_n_cnt*(i+1)] for i in range(n_cpu)]\ndfs[-1] = X_train.iloc[train_n_cnt*(n_cpu-1):]\nres = pool.map(preprocess_data, [x_df for x_df in dfs])\npool.close()\n\n# Pre-procecss validation set\npool = mp.Pool(n_cpu)\ndfs = [X_val.iloc[val_n_cnt*i:val_n_cnt*(i+1)] for i in range(n_cpu)]\ndfs[-1] = X_val.iloc[val_n_cnt*(n_cpu-1):] \nres = pool.map(preprocess_data, [x_df for x_df in dfs])\npool.close()\n\n# Pre-procecss test set\npool = mp.Pool(n_cpu)\ndfs = [test.iloc[test_n_cnt*i:test_n_cnt*(i+1)] for i in range(n_cpu)]\ndfs[-1] = test.iloc[test_n_cnt*(n_cpu-1):] \nres = pool.map(preprocess_test, [x_df for x_df in dfs])\npool.close()", "_____no_output_____" ] ], [ [ "# Data generator", "_____no_output_____" ] ], [ [ "datagen=ImageDataGenerator(rescale=1./255, \n rotation_range=360,\n horizontal_flip=True,\n vertical_flip=True)\n\ntrain_generator=datagen.flow_from_dataframe(\n dataframe=X_train,\n directory=train_dest_path,\n x_col=\"id_code\",\n y_col=\"diagnosis\",\n class_mode=\"raw\",\n batch_size=BATCH_SIZE,\n target_size=(HEIGHT, WIDTH),\n seed=seed)\n\nvalid_generator=datagen.flow_from_dataframe(\n dataframe=X_val,\n directory=validation_dest_path,\n x_col=\"id_code\",\n y_col=\"diagnosis\",\n class_mode=\"raw\",\n batch_size=BATCH_SIZE,\n target_size=(HEIGHT, WIDTH),\n seed=seed)\n\ntest_generator=datagen.flow_from_dataframe( \n dataframe=test,\n directory=test_dest_path,\n x_col=\"id_code\",\n batch_size=1,\n class_mode=None,\n shuffle=False,\n target_size=(HEIGHT, WIDTH),\n seed=seed)", "Found 18697 validated image filenames.\nFound 733 validated image filenames.\nFound 1928 validated image filenames.\n" ], [ "def classify(x):\n if x < 0.5:\n return 0\n elif x < 1.5:\n return 1\n elif x < 2.5:\n return 2\n elif x < 3.5:\n return 3\n return 4\n\nlabels = ['0 - No DR', '1 - Mild', '2 - Moderate', '3 - Severe', '4 - Proliferative DR']\ndef plot_confusion_matrix(train, validation, labels=labels):\n train_labels, train_preds = train\n validation_labels, validation_preds = validation\n fig, (ax1, ax2) = plt.subplots(1, 2, sharex='col', figsize=(24, 7))\n train_cnf_matrix = confusion_matrix(train_labels, train_preds)\n validation_cnf_matrix = confusion_matrix(validation_labels, validation_preds)\n\n train_cnf_matrix_norm = train_cnf_matrix.astype('float') / train_cnf_matrix.sum(axis=1)[:, np.newaxis]\n validation_cnf_matrix_norm = validation_cnf_matrix.astype('float') / validation_cnf_matrix.sum(axis=1)[:, np.newaxis]\n\n train_df_cm = pd.DataFrame(train_cnf_matrix_norm, index=labels, columns=labels)\n validation_df_cm = pd.DataFrame(validation_cnf_matrix_norm, index=labels, columns=labels)\n\n sns.heatmap(train_df_cm, annot=True, fmt='.2f', cmap=\"Blues\",ax=ax1).set_title('Train')\n sns.heatmap(validation_df_cm, annot=True, fmt='.2f', cmap=sns.cubehelix_palette(8),ax=ax2).set_title('Validation')\n plt.show()\n \ndef plot_metrics(history, figsize=(20, 14)):\n fig, (ax1, ax2) = plt.subplots(2, 1, sharex='col', figsize=figsize)\n\n ax1.plot(history['loss'], label='Train loss')\n ax1.plot(history['val_loss'], label='Validation loss')\n ax1.legend(loc='best')\n ax1.set_title('Loss')\n\n ax2.plot(history['acc'], label='Train accuracy')\n ax2.plot(history['val_acc'], label='Validation accuracy')\n ax2.legend(loc='best')\n ax2.set_title('Accuracy')\n\n plt.xlabel('Epochs')\n sns.despine()\n plt.show()\n \ndef apply_tta(model, generator, steps=10):\n step_size = generator.n//generator.batch_size\n preds_tta = []\n for i in range(steps):\n generator.reset()\n preds = model.predict_generator(generator, steps=step_size)\n preds_tta.append(preds)\n\n return np.mean(preds_tta, axis=0)\n\ndef evaluate_model(train, validation):\n train_labels, train_preds = train\n validation_labels, validation_preds = validation\n print(\"Train Cohen Kappa score: %.3f\" % cohen_kappa_score(train_preds, train_labels, weights='quadratic'))\n print(\"Validation Cohen Kappa score: %.3f\" % cohen_kappa_score(validation_preds, validation_labels, weights='quadratic'))\n print(\"Complete set Cohen Kappa score: %.3f\" % cohen_kappa_score(np.append(train_preds, validation_preds), np.append(train_labels, validation_labels), weights='quadratic'))\n\ndef cosine_decay_with_warmup(global_step,\n learning_rate_base,\n total_steps,\n warmup_learning_rate=0.0,\n warmup_steps=0,\n hold_base_rate_steps=0):\n \"\"\"\n Cosine decay schedule with warm up period.\n In this schedule, the learning rate grows linearly from warmup_learning_rate\n to learning_rate_base for warmup_steps, then transitions to a cosine decay\n schedule.\n :param global_step {int}: global step.\n :param learning_rate_base {float}: base learning rate.\n :param total_steps {int}: total number of training steps.\n :param warmup_learning_rate {float}: initial learning rate for warm up. (default: {0.0}).\n :param warmup_steps {int}: number of warmup steps. (default: {0}).\n :param hold_base_rate_steps {int}: Optional number of steps to hold base learning rate before decaying. (default: {0}).\n :param global_step {int}: global step.\n :Returns : a float representing learning rate.\n :Raises ValueError: if warmup_learning_rate is larger than learning_rate_base, or if warmup_steps is larger than total_steps.\n \"\"\"\n\n if total_steps < warmup_steps:\n raise ValueError('total_steps must be larger or equal to warmup_steps.')\n learning_rate = 0.5 * learning_rate_base * (1 + np.cos(\n np.pi *\n (global_step - warmup_steps - hold_base_rate_steps\n ) / float(total_steps - warmup_steps - hold_base_rate_steps)))\n if hold_base_rate_steps > 0:\n learning_rate = np.where(global_step > warmup_steps + hold_base_rate_steps,\n learning_rate, learning_rate_base)\n if warmup_steps > 0:\n if learning_rate_base < warmup_learning_rate:\n raise ValueError('learning_rate_base must be larger or equal to warmup_learning_rate.')\n slope = (learning_rate_base - warmup_learning_rate) / warmup_steps\n warmup_rate = slope * global_step + warmup_learning_rate\n learning_rate = np.where(global_step < warmup_steps, warmup_rate,\n learning_rate)\n return np.where(global_step > total_steps, 0.0, learning_rate)\n\n\nclass WarmUpCosineDecayScheduler(Callback):\n \"\"\"Cosine decay with warmup learning rate scheduler\"\"\"\n\n def __init__(self,\n learning_rate_base,\n total_steps,\n global_step_init=0,\n warmup_learning_rate=0.0,\n warmup_steps=0,\n hold_base_rate_steps=0,\n verbose=0):\n \"\"\"\n Constructor for cosine decay with warmup learning rate scheduler.\n :param learning_rate_base {float}: base learning rate.\n :param total_steps {int}: total number of training steps.\n :param global_step_init {int}: initial global step, e.g. from previous checkpoint.\n :param warmup_learning_rate {float}: initial learning rate for warm up. (default: {0.0}).\n :param warmup_steps {int}: number of warmup steps. (default: {0}).\n :param hold_base_rate_steps {int}: Optional number of steps to hold base learning rate before decaying. (default: {0}).\n :param verbose {int}: quiet, 1: update messages. (default: {0}).\n \"\"\"\n\n super(WarmUpCosineDecayScheduler, self).__init__()\n self.learning_rate_base = learning_rate_base\n self.total_steps = total_steps\n self.global_step = global_step_init\n self.warmup_learning_rate = warmup_learning_rate\n self.warmup_steps = warmup_steps\n self.hold_base_rate_steps = hold_base_rate_steps\n self.verbose = verbose\n self.learning_rates = []\n\n def on_batch_end(self, batch, logs=None):\n self.global_step = self.global_step + 1\n lr = K.get_value(self.model.optimizer.lr)\n self.learning_rates.append(lr)\n\n def on_batch_begin(self, batch, logs=None):\n lr = cosine_decay_with_warmup(global_step=self.global_step,\n learning_rate_base=self.learning_rate_base,\n total_steps=self.total_steps,\n warmup_learning_rate=self.warmup_learning_rate,\n warmup_steps=self.warmup_steps,\n hold_base_rate_steps=self.hold_base_rate_steps)\n K.set_value(self.model.optimizer.lr, lr)\n if self.verbose > 0:\n print('\\nBatch %02d: setting learning rate to %s.' % (self.global_step + 1, lr))", "_____no_output_____" ] ], [ [ "# Model", "_____no_output_____" ] ], [ [ "def create_model(input_shape):\n input_tensor = Input(shape=input_shape)\n base_model = EfficientNetB4(weights=None, \n include_top=False,\n input_tensor=input_tensor)\n base_model.load_weights('../input/efficientnet-keras-weights-b0b5/efficientnet-b4_imagenet_1000_notop.h5')\n\n x = GlobalAveragePooling2D()(base_model.output)\n final_output = Dense(1, activation='linear', name='final_output')(x)\n model = Model(input_tensor, final_output)\n \n return model", "_____no_output_____" ] ], [ [ "# Train top layers", "_____no_output_____" ] ], [ [ "model = create_model(input_shape=(HEIGHT, WIDTH, CHANNELS))\n\nfor layer in model.layers:\n layer.trainable = False\n\nfor i in range(-2, 0):\n model.layers[i].trainable = True\n\nmetric_list = [\"accuracy\"]\noptimizer = optimizers.Adam(lr=WARMUP_LEARNING_RATE)\nmodel.compile(optimizer=optimizer, loss='mean_squared_error', metrics=metric_list)\nmodel.summary()", "__________________________________________________________________________________________________\nLayer (type) Output Shape Param # Connected to \n==================================================================================================\ninput_1 (InputLayer) (None, 256, 256, 3) 0 \n__________________________________________________________________________________________________\nconv2d_1 (Conv2D) (None, 128, 128, 48) 1296 input_1[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_1 (BatchNor (None, 128, 128, 48) 192 conv2d_1[0][0] \n__________________________________________________________________________________________________\nswish_1 (Swish) (None, 128, 128, 48) 0 batch_normalization_1[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_1 (DepthwiseCo (None, 128, 128, 48) 432 swish_1[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_2 (BatchNor (None, 128, 128, 48) 192 depthwise_conv2d_1[0][0] \n__________________________________________________________________________________________________\nswish_2 (Swish) (None, 128, 128, 48) 0 batch_normalization_2[0][0] \n__________________________________________________________________________________________________\nlambda_1 (Lambda) (None, 1, 1, 48) 0 swish_2[0][0] \n__________________________________________________________________________________________________\nconv2d_2 (Conv2D) (None, 1, 1, 12) 588 lambda_1[0][0] \n__________________________________________________________________________________________________\nswish_3 (Swish) (None, 1, 1, 12) 0 conv2d_2[0][0] \n__________________________________________________________________________________________________\nconv2d_3 (Conv2D) (None, 1, 1, 48) 624 swish_3[0][0] \n__________________________________________________________________________________________________\nactivation_1 (Activation) (None, 1, 1, 48) 0 conv2d_3[0][0] \n__________________________________________________________________________________________________\nmultiply_1 (Multiply) (None, 128, 128, 48) 0 activation_1[0][0] \n swish_2[0][0] \n__________________________________________________________________________________________________\nconv2d_4 (Conv2D) (None, 128, 128, 24) 1152 multiply_1[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_3 (BatchNor (None, 128, 128, 24) 96 conv2d_4[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_2 (DepthwiseCo (None, 128, 128, 24) 216 batch_normalization_3[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_4 (BatchNor (None, 128, 128, 24) 96 depthwise_conv2d_2[0][0] \n__________________________________________________________________________________________________\nswish_4 (Swish) (None, 128, 128, 24) 0 batch_normalization_4[0][0] \n__________________________________________________________________________________________________\nlambda_2 (Lambda) (None, 1, 1, 24) 0 swish_4[0][0] \n__________________________________________________________________________________________________\nconv2d_5 (Conv2D) (None, 1, 1, 6) 150 lambda_2[0][0] \n__________________________________________________________________________________________________\nswish_5 (Swish) (None, 1, 1, 6) 0 conv2d_5[0][0] \n__________________________________________________________________________________________________\nconv2d_6 (Conv2D) (None, 1, 1, 24) 168 swish_5[0][0] \n__________________________________________________________________________________________________\nactivation_2 (Activation) (None, 1, 1, 24) 0 conv2d_6[0][0] \n__________________________________________________________________________________________________\nmultiply_2 (Multiply) (None, 128, 128, 24) 0 activation_2[0][0] \n swish_4[0][0] \n__________________________________________________________________________________________________\nconv2d_7 (Conv2D) (None, 128, 128, 24) 576 multiply_2[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_5 (BatchNor (None, 128, 128, 24) 96 conv2d_7[0][0] \n__________________________________________________________________________________________________\ndrop_connect_1 (DropConnect) (None, 128, 128, 24) 0 batch_normalization_5[0][0] \n__________________________________________________________________________________________________\nadd_1 (Add) (None, 128, 128, 24) 0 drop_connect_1[0][0] \n batch_normalization_3[0][0] \n__________________________________________________________________________________________________\nconv2d_8 (Conv2D) (None, 128, 128, 144 3456 add_1[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_6 (BatchNor (None, 128, 128, 144 576 conv2d_8[0][0] \n__________________________________________________________________________________________________\nswish_6 (Swish) (None, 128, 128, 144 0 batch_normalization_6[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_3 (DepthwiseCo (None, 64, 64, 144) 1296 swish_6[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_7 (BatchNor (None, 64, 64, 144) 576 depthwise_conv2d_3[0][0] \n__________________________________________________________________________________________________\nswish_7 (Swish) (None, 64, 64, 144) 0 batch_normalization_7[0][0] \n__________________________________________________________________________________________________\nlambda_3 (Lambda) (None, 1, 1, 144) 0 swish_7[0][0] \n__________________________________________________________________________________________________\nconv2d_9 (Conv2D) (None, 1, 1, 6) 870 lambda_3[0][0] \n__________________________________________________________________________________________________\nswish_8 (Swish) (None, 1, 1, 6) 0 conv2d_9[0][0] \n__________________________________________________________________________________________________\nconv2d_10 (Conv2D) (None, 1, 1, 144) 1008 swish_8[0][0] \n__________________________________________________________________________________________________\nactivation_3 (Activation) (None, 1, 1, 144) 0 conv2d_10[0][0] \n__________________________________________________________________________________________________\nmultiply_3 (Multiply) (None, 64, 64, 144) 0 activation_3[0][0] \n swish_7[0][0] \n__________________________________________________________________________________________________\nconv2d_11 (Conv2D) (None, 64, 64, 32) 4608 multiply_3[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_8 (BatchNor (None, 64, 64, 32) 128 conv2d_11[0][0] \n__________________________________________________________________________________________________\nconv2d_12 (Conv2D) (None, 64, 64, 192) 6144 batch_normalization_8[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_9 (BatchNor (None, 64, 64, 192) 768 conv2d_12[0][0] \n__________________________________________________________________________________________________\nswish_9 (Swish) (None, 64, 64, 192) 0 batch_normalization_9[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_4 (DepthwiseCo (None, 64, 64, 192) 1728 swish_9[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_10 (BatchNo (None, 64, 64, 192) 768 depthwise_conv2d_4[0][0] \n__________________________________________________________________________________________________\nswish_10 (Swish) (None, 64, 64, 192) 0 batch_normalization_10[0][0] \n__________________________________________________________________________________________________\nlambda_4 (Lambda) (None, 1, 1, 192) 0 swish_10[0][0] \n__________________________________________________________________________________________________\nconv2d_13 (Conv2D) (None, 1, 1, 8) 1544 lambda_4[0][0] \n__________________________________________________________________________________________________\nswish_11 (Swish) (None, 1, 1, 8) 0 conv2d_13[0][0] \n__________________________________________________________________________________________________\nconv2d_14 (Conv2D) (None, 1, 1, 192) 1728 swish_11[0][0] \n__________________________________________________________________________________________________\nactivation_4 (Activation) (None, 1, 1, 192) 0 conv2d_14[0][0] \n__________________________________________________________________________________________________\nmultiply_4 (Multiply) (None, 64, 64, 192) 0 activation_4[0][0] \n swish_10[0][0] \n__________________________________________________________________________________________________\nconv2d_15 (Conv2D) (None, 64, 64, 32) 6144 multiply_4[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_11 (BatchNo (None, 64, 64, 32) 128 conv2d_15[0][0] \n__________________________________________________________________________________________________\ndrop_connect_2 (DropConnect) (None, 64, 64, 32) 0 batch_normalization_11[0][0] \n__________________________________________________________________________________________________\nadd_2 (Add) (None, 64, 64, 32) 0 drop_connect_2[0][0] \n batch_normalization_8[0][0] \n__________________________________________________________________________________________________\nconv2d_16 (Conv2D) (None, 64, 64, 192) 6144 add_2[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_12 (BatchNo (None, 64, 64, 192) 768 conv2d_16[0][0] \n__________________________________________________________________________________________________\nswish_12 (Swish) (None, 64, 64, 192) 0 batch_normalization_12[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_5 (DepthwiseCo (None, 64, 64, 192) 1728 swish_12[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_13 (BatchNo (None, 64, 64, 192) 768 depthwise_conv2d_5[0][0] \n__________________________________________________________________________________________________\nswish_13 (Swish) (None, 64, 64, 192) 0 batch_normalization_13[0][0] \n__________________________________________________________________________________________________\nlambda_5 (Lambda) (None, 1, 1, 192) 0 swish_13[0][0] \n__________________________________________________________________________________________________\nconv2d_17 (Conv2D) (None, 1, 1, 8) 1544 lambda_5[0][0] \n__________________________________________________________________________________________________\nswish_14 (Swish) (None, 1, 1, 8) 0 conv2d_17[0][0] \n__________________________________________________________________________________________________\nconv2d_18 (Conv2D) (None, 1, 1, 192) 1728 swish_14[0][0] \n__________________________________________________________________________________________________\nactivation_5 (Activation) (None, 1, 1, 192) 0 conv2d_18[0][0] \n__________________________________________________________________________________________________\nmultiply_5 (Multiply) (None, 64, 64, 192) 0 activation_5[0][0] \n swish_13[0][0] \n__________________________________________________________________________________________________\nconv2d_19 (Conv2D) (None, 64, 64, 32) 6144 multiply_5[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_14 (BatchNo (None, 64, 64, 32) 128 conv2d_19[0][0] \n__________________________________________________________________________________________________\ndrop_connect_3 (DropConnect) (None, 64, 64, 32) 0 batch_normalization_14[0][0] \n__________________________________________________________________________________________________\nadd_3 (Add) (None, 64, 64, 32) 0 drop_connect_3[0][0] \n add_2[0][0] \n__________________________________________________________________________________________________\nconv2d_20 (Conv2D) (None, 64, 64, 192) 6144 add_3[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_15 (BatchNo (None, 64, 64, 192) 768 conv2d_20[0][0] \n__________________________________________________________________________________________________\nswish_15 (Swish) (None, 64, 64, 192) 0 batch_normalization_15[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_6 (DepthwiseCo (None, 64, 64, 192) 1728 swish_15[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_16 (BatchNo (None, 64, 64, 192) 768 depthwise_conv2d_6[0][0] \n__________________________________________________________________________________________________\nswish_16 (Swish) (None, 64, 64, 192) 0 batch_normalization_16[0][0] \n__________________________________________________________________________________________________\nlambda_6 (Lambda) (None, 1, 1, 192) 0 swish_16[0][0] \n__________________________________________________________________________________________________\nconv2d_21 (Conv2D) (None, 1, 1, 8) 1544 lambda_6[0][0] \n__________________________________________________________________________________________________\nswish_17 (Swish) (None, 1, 1, 8) 0 conv2d_21[0][0] \n__________________________________________________________________________________________________\nconv2d_22 (Conv2D) (None, 1, 1, 192) 1728 swish_17[0][0] \n__________________________________________________________________________________________________\nactivation_6 (Activation) (None, 1, 1, 192) 0 conv2d_22[0][0] \n__________________________________________________________________________________________________\nmultiply_6 (Multiply) (None, 64, 64, 192) 0 activation_6[0][0] \n swish_16[0][0] \n__________________________________________________________________________________________________\nconv2d_23 (Conv2D) (None, 64, 64, 32) 6144 multiply_6[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_17 (BatchNo (None, 64, 64, 32) 128 conv2d_23[0][0] \n__________________________________________________________________________________________________\ndrop_connect_4 (DropConnect) (None, 64, 64, 32) 0 batch_normalization_17[0][0] \n__________________________________________________________________________________________________\nadd_4 (Add) (None, 64, 64, 32) 0 drop_connect_4[0][0] \n add_3[0][0] \n__________________________________________________________________________________________________\nconv2d_24 (Conv2D) (None, 64, 64, 192) 6144 add_4[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_18 (BatchNo (None, 64, 64, 192) 768 conv2d_24[0][0] \n__________________________________________________________________________________________________\nswish_18 (Swish) (None, 64, 64, 192) 0 batch_normalization_18[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_7 (DepthwiseCo (None, 32, 32, 192) 4800 swish_18[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_19 (BatchNo (None, 32, 32, 192) 768 depthwise_conv2d_7[0][0] \n__________________________________________________________________________________________________\nswish_19 (Swish) (None, 32, 32, 192) 0 batch_normalization_19[0][0] \n__________________________________________________________________________________________________\nlambda_7 (Lambda) (None, 1, 1, 192) 0 swish_19[0][0] \n__________________________________________________________________________________________________\nconv2d_25 (Conv2D) (None, 1, 1, 8) 1544 lambda_7[0][0] \n__________________________________________________________________________________________________\nswish_20 (Swish) (None, 1, 1, 8) 0 conv2d_25[0][0] \n__________________________________________________________________________________________________\nconv2d_26 (Conv2D) (None, 1, 1, 192) 1728 swish_20[0][0] \n__________________________________________________________________________________________________\nactivation_7 (Activation) (None, 1, 1, 192) 0 conv2d_26[0][0] \n__________________________________________________________________________________________________\nmultiply_7 (Multiply) (None, 32, 32, 192) 0 activation_7[0][0] \n swish_19[0][0] \n__________________________________________________________________________________________________\nconv2d_27 (Conv2D) (None, 32, 32, 56) 10752 multiply_7[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_20 (BatchNo (None, 32, 32, 56) 224 conv2d_27[0][0] \n__________________________________________________________________________________________________\nconv2d_28 (Conv2D) (None, 32, 32, 336) 18816 batch_normalization_20[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_21 (BatchNo (None, 32, 32, 336) 1344 conv2d_28[0][0] \n__________________________________________________________________________________________________\nswish_21 (Swish) (None, 32, 32, 336) 0 batch_normalization_21[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_8 (DepthwiseCo (None, 32, 32, 336) 8400 swish_21[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_22 (BatchNo (None, 32, 32, 336) 1344 depthwise_conv2d_8[0][0] \n__________________________________________________________________________________________________\nswish_22 (Swish) (None, 32, 32, 336) 0 batch_normalization_22[0][0] \n__________________________________________________________________________________________________\nlambda_8 (Lambda) (None, 1, 1, 336) 0 swish_22[0][0] \n__________________________________________________________________________________________________\nconv2d_29 (Conv2D) (None, 1, 1, 14) 4718 lambda_8[0][0] \n__________________________________________________________________________________________________\nswish_23 (Swish) (None, 1, 1, 14) 0 conv2d_29[0][0] \n__________________________________________________________________________________________________\nconv2d_30 (Conv2D) (None, 1, 1, 336) 5040 swish_23[0][0] \n__________________________________________________________________________________________________\nactivation_8 (Activation) (None, 1, 1, 336) 0 conv2d_30[0][0] \n__________________________________________________________________________________________________\nmultiply_8 (Multiply) (None, 32, 32, 336) 0 activation_8[0][0] \n swish_22[0][0] \n__________________________________________________________________________________________________\nconv2d_31 (Conv2D) (None, 32, 32, 56) 18816 multiply_8[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_23 (BatchNo (None, 32, 32, 56) 224 conv2d_31[0][0] \n__________________________________________________________________________________________________\ndrop_connect_5 (DropConnect) (None, 32, 32, 56) 0 batch_normalization_23[0][0] \n__________________________________________________________________________________________________\nadd_5 (Add) (None, 32, 32, 56) 0 drop_connect_5[0][0] \n batch_normalization_20[0][0] \n__________________________________________________________________________________________________\nconv2d_32 (Conv2D) (None, 32, 32, 336) 18816 add_5[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_24 (BatchNo (None, 32, 32, 336) 1344 conv2d_32[0][0] \n__________________________________________________________________________________________________\nswish_24 (Swish) (None, 32, 32, 336) 0 batch_normalization_24[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_9 (DepthwiseCo (None, 32, 32, 336) 8400 swish_24[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_25 (BatchNo (None, 32, 32, 336) 1344 depthwise_conv2d_9[0][0] \n__________________________________________________________________________________________________\nswish_25 (Swish) (None, 32, 32, 336) 0 batch_normalization_25[0][0] \n__________________________________________________________________________________________________\nlambda_9 (Lambda) (None, 1, 1, 336) 0 swish_25[0][0] \n__________________________________________________________________________________________________\nconv2d_33 (Conv2D) (None, 1, 1, 14) 4718 lambda_9[0][0] \n__________________________________________________________________________________________________\nswish_26 (Swish) (None, 1, 1, 14) 0 conv2d_33[0][0] \n__________________________________________________________________________________________________\nconv2d_34 (Conv2D) (None, 1, 1, 336) 5040 swish_26[0][0] \n__________________________________________________________________________________________________\nactivation_9 (Activation) (None, 1, 1, 336) 0 conv2d_34[0][0] \n__________________________________________________________________________________________________\nmultiply_9 (Multiply) (None, 32, 32, 336) 0 activation_9[0][0] \n swish_25[0][0] \n__________________________________________________________________________________________________\nconv2d_35 (Conv2D) (None, 32, 32, 56) 18816 multiply_9[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_26 (BatchNo (None, 32, 32, 56) 224 conv2d_35[0][0] \n__________________________________________________________________________________________________\ndrop_connect_6 (DropConnect) (None, 32, 32, 56) 0 batch_normalization_26[0][0] \n__________________________________________________________________________________________________\nadd_6 (Add) (None, 32, 32, 56) 0 drop_connect_6[0][0] \n add_5[0][0] \n__________________________________________________________________________________________________\nconv2d_36 (Conv2D) (None, 32, 32, 336) 18816 add_6[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_27 (BatchNo (None, 32, 32, 336) 1344 conv2d_36[0][0] \n__________________________________________________________________________________________________\nswish_27 (Swish) (None, 32, 32, 336) 0 batch_normalization_27[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_10 (DepthwiseC (None, 32, 32, 336) 8400 swish_27[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_28 (BatchNo (None, 32, 32, 336) 1344 depthwise_conv2d_10[0][0] \n__________________________________________________________________________________________________\nswish_28 (Swish) (None, 32, 32, 336) 0 batch_normalization_28[0][0] \n__________________________________________________________________________________________________\nlambda_10 (Lambda) (None, 1, 1, 336) 0 swish_28[0][0] \n__________________________________________________________________________________________________\nconv2d_37 (Conv2D) (None, 1, 1, 14) 4718 lambda_10[0][0] \n__________________________________________________________________________________________________\nswish_29 (Swish) (None, 1, 1, 14) 0 conv2d_37[0][0] \n__________________________________________________________________________________________________\nconv2d_38 (Conv2D) (None, 1, 1, 336) 5040 swish_29[0][0] \n__________________________________________________________________________________________________\nactivation_10 (Activation) (None, 1, 1, 336) 0 conv2d_38[0][0] \n__________________________________________________________________________________________________\nmultiply_10 (Multiply) (None, 32, 32, 336) 0 activation_10[0][0] \n swish_28[0][0] \n__________________________________________________________________________________________________\nconv2d_39 (Conv2D) (None, 32, 32, 56) 18816 multiply_10[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_29 (BatchNo (None, 32, 32, 56) 224 conv2d_39[0][0] \n__________________________________________________________________________________________________\ndrop_connect_7 (DropConnect) (None, 32, 32, 56) 0 batch_normalization_29[0][0] \n__________________________________________________________________________________________________\nadd_7 (Add) (None, 32, 32, 56) 0 drop_connect_7[0][0] \n add_6[0][0] \n__________________________________________________________________________________________________\nconv2d_40 (Conv2D) (None, 32, 32, 336) 18816 add_7[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_30 (BatchNo (None, 32, 32, 336) 1344 conv2d_40[0][0] \n__________________________________________________________________________________________________\nswish_30 (Swish) (None, 32, 32, 336) 0 batch_normalization_30[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_11 (DepthwiseC (None, 16, 16, 336) 3024 swish_30[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_31 (BatchNo (None, 16, 16, 336) 1344 depthwise_conv2d_11[0][0] \n__________________________________________________________________________________________________\nswish_31 (Swish) (None, 16, 16, 336) 0 batch_normalization_31[0][0] \n__________________________________________________________________________________________________\nlambda_11 (Lambda) (None, 1, 1, 336) 0 swish_31[0][0] \n__________________________________________________________________________________________________\nconv2d_41 (Conv2D) (None, 1, 1, 14) 4718 lambda_11[0][0] \n__________________________________________________________________________________________________\nswish_32 (Swish) (None, 1, 1, 14) 0 conv2d_41[0][0] \n__________________________________________________________________________________________________\nconv2d_42 (Conv2D) (None, 1, 1, 336) 5040 swish_32[0][0] \n__________________________________________________________________________________________________\nactivation_11 (Activation) (None, 1, 1, 336) 0 conv2d_42[0][0] \n__________________________________________________________________________________________________\nmultiply_11 (Multiply) (None, 16, 16, 336) 0 activation_11[0][0] \n swish_31[0][0] \n__________________________________________________________________________________________________\nconv2d_43 (Conv2D) (None, 16, 16, 112) 37632 multiply_11[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_32 (BatchNo (None, 16, 16, 112) 448 conv2d_43[0][0] \n__________________________________________________________________________________________________\nconv2d_44 (Conv2D) (None, 16, 16, 672) 75264 batch_normalization_32[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_33 (BatchNo (None, 16, 16, 672) 2688 conv2d_44[0][0] \n__________________________________________________________________________________________________\nswish_33 (Swish) (None, 16, 16, 672) 0 batch_normalization_33[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_12 (DepthwiseC (None, 16, 16, 672) 6048 swish_33[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_34 (BatchNo (None, 16, 16, 672) 2688 depthwise_conv2d_12[0][0] \n__________________________________________________________________________________________________\nswish_34 (Swish) (None, 16, 16, 672) 0 batch_normalization_34[0][0] \n__________________________________________________________________________________________________\nlambda_12 (Lambda) (None, 1, 1, 672) 0 swish_34[0][0] \n__________________________________________________________________________________________________\nconv2d_45 (Conv2D) (None, 1, 1, 28) 18844 lambda_12[0][0] \n__________________________________________________________________________________________________\nswish_35 (Swish) (None, 1, 1, 28) 0 conv2d_45[0][0] \n__________________________________________________________________________________________________\nconv2d_46 (Conv2D) (None, 1, 1, 672) 19488 swish_35[0][0] \n__________________________________________________________________________________________________\nactivation_12 (Activation) (None, 1, 1, 672) 0 conv2d_46[0][0] \n__________________________________________________________________________________________________\nmultiply_12 (Multiply) (None, 16, 16, 672) 0 activation_12[0][0] \n swish_34[0][0] \n__________________________________________________________________________________________________\nconv2d_47 (Conv2D) (None, 16, 16, 112) 75264 multiply_12[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_35 (BatchNo (None, 16, 16, 112) 448 conv2d_47[0][0] \n__________________________________________________________________________________________________\ndrop_connect_8 (DropConnect) (None, 16, 16, 112) 0 batch_normalization_35[0][0] \n__________________________________________________________________________________________________\nadd_8 (Add) (None, 16, 16, 112) 0 drop_connect_8[0][0] \n batch_normalization_32[0][0] \n__________________________________________________________________________________________________\nconv2d_48 (Conv2D) (None, 16, 16, 672) 75264 add_8[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_36 (BatchNo (None, 16, 16, 672) 2688 conv2d_48[0][0] \n__________________________________________________________________________________________________\nswish_36 (Swish) (None, 16, 16, 672) 0 batch_normalization_36[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_13 (DepthwiseC (None, 16, 16, 672) 6048 swish_36[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_37 (BatchNo (None, 16, 16, 672) 2688 depthwise_conv2d_13[0][0] \n__________________________________________________________________________________________________\nswish_37 (Swish) (None, 16, 16, 672) 0 batch_normalization_37[0][0] \n__________________________________________________________________________________________________\nlambda_13 (Lambda) (None, 1, 1, 672) 0 swish_37[0][0] \n__________________________________________________________________________________________________\nconv2d_49 (Conv2D) (None, 1, 1, 28) 18844 lambda_13[0][0] \n__________________________________________________________________________________________________\nswish_38 (Swish) (None, 1, 1, 28) 0 conv2d_49[0][0] \n__________________________________________________________________________________________________\nconv2d_50 (Conv2D) (None, 1, 1, 672) 19488 swish_38[0][0] \n__________________________________________________________________________________________________\nactivation_13 (Activation) (None, 1, 1, 672) 0 conv2d_50[0][0] \n__________________________________________________________________________________________________\nmultiply_13 (Multiply) (None, 16, 16, 672) 0 activation_13[0][0] \n swish_37[0][0] \n__________________________________________________________________________________________________\nconv2d_51 (Conv2D) (None, 16, 16, 112) 75264 multiply_13[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_38 (BatchNo (None, 16, 16, 112) 448 conv2d_51[0][0] \n__________________________________________________________________________________________________\ndrop_connect_9 (DropConnect) (None, 16, 16, 112) 0 batch_normalization_38[0][0] \n__________________________________________________________________________________________________\nadd_9 (Add) (None, 16, 16, 112) 0 drop_connect_9[0][0] \n add_8[0][0] \n__________________________________________________________________________________________________\nconv2d_52 (Conv2D) (None, 16, 16, 672) 75264 add_9[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_39 (BatchNo (None, 16, 16, 672) 2688 conv2d_52[0][0] \n__________________________________________________________________________________________________\nswish_39 (Swish) (None, 16, 16, 672) 0 batch_normalization_39[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_14 (DepthwiseC (None, 16, 16, 672) 6048 swish_39[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_40 (BatchNo (None, 16, 16, 672) 2688 depthwise_conv2d_14[0][0] \n__________________________________________________________________________________________________\nswish_40 (Swish) (None, 16, 16, 672) 0 batch_normalization_40[0][0] \n__________________________________________________________________________________________________\nlambda_14 (Lambda) (None, 1, 1, 672) 0 swish_40[0][0] \n__________________________________________________________________________________________________\nconv2d_53 (Conv2D) (None, 1, 1, 28) 18844 lambda_14[0][0] \n__________________________________________________________________________________________________\nswish_41 (Swish) (None, 1, 1, 28) 0 conv2d_53[0][0] \n__________________________________________________________________________________________________\nconv2d_54 (Conv2D) (None, 1, 1, 672) 19488 swish_41[0][0] \n__________________________________________________________________________________________________\nactivation_14 (Activation) (None, 1, 1, 672) 0 conv2d_54[0][0] \n__________________________________________________________________________________________________\nmultiply_14 (Multiply) (None, 16, 16, 672) 0 activation_14[0][0] \n swish_40[0][0] \n__________________________________________________________________________________________________\nconv2d_55 (Conv2D) (None, 16, 16, 112) 75264 multiply_14[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_41 (BatchNo (None, 16, 16, 112) 448 conv2d_55[0][0] \n__________________________________________________________________________________________________\ndrop_connect_10 (DropConnect) (None, 16, 16, 112) 0 batch_normalization_41[0][0] \n__________________________________________________________________________________________________\nadd_10 (Add) (None, 16, 16, 112) 0 drop_connect_10[0][0] \n add_9[0][0] \n__________________________________________________________________________________________________\nconv2d_56 (Conv2D) (None, 16, 16, 672) 75264 add_10[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_42 (BatchNo (None, 16, 16, 672) 2688 conv2d_56[0][0] \n__________________________________________________________________________________________________\nswish_42 (Swish) (None, 16, 16, 672) 0 batch_normalization_42[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_15 (DepthwiseC (None, 16, 16, 672) 6048 swish_42[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_43 (BatchNo (None, 16, 16, 672) 2688 depthwise_conv2d_15[0][0] \n__________________________________________________________________________________________________\nswish_43 (Swish) (None, 16, 16, 672) 0 batch_normalization_43[0][0] \n__________________________________________________________________________________________________\nlambda_15 (Lambda) (None, 1, 1, 672) 0 swish_43[0][0] \n__________________________________________________________________________________________________\nconv2d_57 (Conv2D) (None, 1, 1, 28) 18844 lambda_15[0][0] \n__________________________________________________________________________________________________\nswish_44 (Swish) (None, 1, 1, 28) 0 conv2d_57[0][0] \n__________________________________________________________________________________________________\nconv2d_58 (Conv2D) (None, 1, 1, 672) 19488 swish_44[0][0] \n__________________________________________________________________________________________________\nactivation_15 (Activation) (None, 1, 1, 672) 0 conv2d_58[0][0] \n__________________________________________________________________________________________________\nmultiply_15 (Multiply) (None, 16, 16, 672) 0 activation_15[0][0] \n swish_43[0][0] \n__________________________________________________________________________________________________\nconv2d_59 (Conv2D) (None, 16, 16, 112) 75264 multiply_15[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_44 (BatchNo (None, 16, 16, 112) 448 conv2d_59[0][0] \n__________________________________________________________________________________________________\ndrop_connect_11 (DropConnect) (None, 16, 16, 112) 0 batch_normalization_44[0][0] \n__________________________________________________________________________________________________\nadd_11 (Add) (None, 16, 16, 112) 0 drop_connect_11[0][0] \n add_10[0][0] \n__________________________________________________________________________________________________\nconv2d_60 (Conv2D) (None, 16, 16, 672) 75264 add_11[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_45 (BatchNo (None, 16, 16, 672) 2688 conv2d_60[0][0] \n__________________________________________________________________________________________________\nswish_45 (Swish) (None, 16, 16, 672) 0 batch_normalization_45[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_16 (DepthwiseC (None, 16, 16, 672) 6048 swish_45[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_46 (BatchNo (None, 16, 16, 672) 2688 depthwise_conv2d_16[0][0] \n__________________________________________________________________________________________________\nswish_46 (Swish) (None, 16, 16, 672) 0 batch_normalization_46[0][0] \n__________________________________________________________________________________________________\nlambda_16 (Lambda) (None, 1, 1, 672) 0 swish_46[0][0] \n__________________________________________________________________________________________________\nconv2d_61 (Conv2D) (None, 1, 1, 28) 18844 lambda_16[0][0] \n__________________________________________________________________________________________________\nswish_47 (Swish) (None, 1, 1, 28) 0 conv2d_61[0][0] \n__________________________________________________________________________________________________\nconv2d_62 (Conv2D) (None, 1, 1, 672) 19488 swish_47[0][0] \n__________________________________________________________________________________________________\nactivation_16 (Activation) (None, 1, 1, 672) 0 conv2d_62[0][0] \n__________________________________________________________________________________________________\nmultiply_16 (Multiply) (None, 16, 16, 672) 0 activation_16[0][0] \n swish_46[0][0] \n__________________________________________________________________________________________________\nconv2d_63 (Conv2D) (None, 16, 16, 112) 75264 multiply_16[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_47 (BatchNo (None, 16, 16, 112) 448 conv2d_63[0][0] \n__________________________________________________________________________________________________\ndrop_connect_12 (DropConnect) (None, 16, 16, 112) 0 batch_normalization_47[0][0] \n__________________________________________________________________________________________________\nadd_12 (Add) (None, 16, 16, 112) 0 drop_connect_12[0][0] \n add_11[0][0] \n__________________________________________________________________________________________________\nconv2d_64 (Conv2D) (None, 16, 16, 672) 75264 add_12[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_48 (BatchNo (None, 16, 16, 672) 2688 conv2d_64[0][0] \n__________________________________________________________________________________________________\nswish_48 (Swish) (None, 16, 16, 672) 0 batch_normalization_48[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_17 (DepthwiseC (None, 16, 16, 672) 16800 swish_48[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_49 (BatchNo (None, 16, 16, 672) 2688 depthwise_conv2d_17[0][0] \n__________________________________________________________________________________________________\nswish_49 (Swish) (None, 16, 16, 672) 0 batch_normalization_49[0][0] \n__________________________________________________________________________________________________\nlambda_17 (Lambda) (None, 1, 1, 672) 0 swish_49[0][0] \n__________________________________________________________________________________________________\nconv2d_65 (Conv2D) (None, 1, 1, 28) 18844 lambda_17[0][0] \n__________________________________________________________________________________________________\nswish_50 (Swish) (None, 1, 1, 28) 0 conv2d_65[0][0] \n__________________________________________________________________________________________________\nconv2d_66 (Conv2D) (None, 1, 1, 672) 19488 swish_50[0][0] \n__________________________________________________________________________________________________\nactivation_17 (Activation) (None, 1, 1, 672) 0 conv2d_66[0][0] \n__________________________________________________________________________________________________\nmultiply_17 (Multiply) (None, 16, 16, 672) 0 activation_17[0][0] \n swish_49[0][0] \n__________________________________________________________________________________________________\nconv2d_67 (Conv2D) (None, 16, 16, 160) 107520 multiply_17[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_50 (BatchNo (None, 16, 16, 160) 640 conv2d_67[0][0] \n__________________________________________________________________________________________________\nconv2d_68 (Conv2D) (None, 16, 16, 960) 153600 batch_normalization_50[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_51 (BatchNo (None, 16, 16, 960) 3840 conv2d_68[0][0] \n__________________________________________________________________________________________________\nswish_51 (Swish) (None, 16, 16, 960) 0 batch_normalization_51[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_18 (DepthwiseC (None, 16, 16, 960) 24000 swish_51[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_52 (BatchNo (None, 16, 16, 960) 3840 depthwise_conv2d_18[0][0] \n__________________________________________________________________________________________________\nswish_52 (Swish) (None, 16, 16, 960) 0 batch_normalization_52[0][0] \n__________________________________________________________________________________________________\nlambda_18 (Lambda) (None, 1, 1, 960) 0 swish_52[0][0] \n__________________________________________________________________________________________________\nconv2d_69 (Conv2D) (None, 1, 1, 40) 38440 lambda_18[0][0] \n__________________________________________________________________________________________________\nswish_53 (Swish) (None, 1, 1, 40) 0 conv2d_69[0][0] \n__________________________________________________________________________________________________\nconv2d_70 (Conv2D) (None, 1, 1, 960) 39360 swish_53[0][0] \n__________________________________________________________________________________________________\nactivation_18 (Activation) (None, 1, 1, 960) 0 conv2d_70[0][0] \n__________________________________________________________________________________________________\nmultiply_18 (Multiply) (None, 16, 16, 960) 0 activation_18[0][0] \n swish_52[0][0] \n__________________________________________________________________________________________________\nconv2d_71 (Conv2D) (None, 16, 16, 160) 153600 multiply_18[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_53 (BatchNo (None, 16, 16, 160) 640 conv2d_71[0][0] \n__________________________________________________________________________________________________\ndrop_connect_13 (DropConnect) (None, 16, 16, 160) 0 batch_normalization_53[0][0] \n__________________________________________________________________________________________________\nadd_13 (Add) (None, 16, 16, 160) 0 drop_connect_13[0][0] \n batch_normalization_50[0][0] \n__________________________________________________________________________________________________\nconv2d_72 (Conv2D) (None, 16, 16, 960) 153600 add_13[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_54 (BatchNo (None, 16, 16, 960) 3840 conv2d_72[0][0] \n__________________________________________________________________________________________________\nswish_54 (Swish) (None, 16, 16, 960) 0 batch_normalization_54[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_19 (DepthwiseC (None, 16, 16, 960) 24000 swish_54[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_55 (BatchNo (None, 16, 16, 960) 3840 depthwise_conv2d_19[0][0] \n__________________________________________________________________________________________________\nswish_55 (Swish) (None, 16, 16, 960) 0 batch_normalization_55[0][0] \n__________________________________________________________________________________________________\nlambda_19 (Lambda) (None, 1, 1, 960) 0 swish_55[0][0] \n__________________________________________________________________________________________________\nconv2d_73 (Conv2D) (None, 1, 1, 40) 38440 lambda_19[0][0] \n__________________________________________________________________________________________________\nswish_56 (Swish) (None, 1, 1, 40) 0 conv2d_73[0][0] \n__________________________________________________________________________________________________\nconv2d_74 (Conv2D) (None, 1, 1, 960) 39360 swish_56[0][0] \n__________________________________________________________________________________________________\nactivation_19 (Activation) (None, 1, 1, 960) 0 conv2d_74[0][0] \n__________________________________________________________________________________________________\nmultiply_19 (Multiply) (None, 16, 16, 960) 0 activation_19[0][0] \n swish_55[0][0] \n__________________________________________________________________________________________________\nconv2d_75 (Conv2D) (None, 16, 16, 160) 153600 multiply_19[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_56 (BatchNo (None, 16, 16, 160) 640 conv2d_75[0][0] \n__________________________________________________________________________________________________\ndrop_connect_14 (DropConnect) (None, 16, 16, 160) 0 batch_normalization_56[0][0] \n__________________________________________________________________________________________________\nadd_14 (Add) (None, 16, 16, 160) 0 drop_connect_14[0][0] \n add_13[0][0] \n__________________________________________________________________________________________________\nconv2d_76 (Conv2D) (None, 16, 16, 960) 153600 add_14[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_57 (BatchNo (None, 16, 16, 960) 3840 conv2d_76[0][0] \n__________________________________________________________________________________________________\nswish_57 (Swish) (None, 16, 16, 960) 0 batch_normalization_57[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_20 (DepthwiseC (None, 16, 16, 960) 24000 swish_57[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_58 (BatchNo (None, 16, 16, 960) 3840 depthwise_conv2d_20[0][0] \n__________________________________________________________________________________________________\nswish_58 (Swish) (None, 16, 16, 960) 0 batch_normalization_58[0][0] \n__________________________________________________________________________________________________\nlambda_20 (Lambda) (None, 1, 1, 960) 0 swish_58[0][0] \n__________________________________________________________________________________________________\nconv2d_77 (Conv2D) (None, 1, 1, 40) 38440 lambda_20[0][0] \n__________________________________________________________________________________________________\nswish_59 (Swish) (None, 1, 1, 40) 0 conv2d_77[0][0] \n__________________________________________________________________________________________________\nconv2d_78 (Conv2D) (None, 1, 1, 960) 39360 swish_59[0][0] \n__________________________________________________________________________________________________\nactivation_20 (Activation) (None, 1, 1, 960) 0 conv2d_78[0][0] \n__________________________________________________________________________________________________\nmultiply_20 (Multiply) (None, 16, 16, 960) 0 activation_20[0][0] \n swish_58[0][0] \n__________________________________________________________________________________________________\nconv2d_79 (Conv2D) (None, 16, 16, 160) 153600 multiply_20[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_59 (BatchNo (None, 16, 16, 160) 640 conv2d_79[0][0] \n__________________________________________________________________________________________________\ndrop_connect_15 (DropConnect) (None, 16, 16, 160) 0 batch_normalization_59[0][0] \n__________________________________________________________________________________________________\nadd_15 (Add) (None, 16, 16, 160) 0 drop_connect_15[0][0] \n add_14[0][0] \n__________________________________________________________________________________________________\nconv2d_80 (Conv2D) (None, 16, 16, 960) 153600 add_15[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_60 (BatchNo (None, 16, 16, 960) 3840 conv2d_80[0][0] \n__________________________________________________________________________________________________\nswish_60 (Swish) (None, 16, 16, 960) 0 batch_normalization_60[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_21 (DepthwiseC (None, 16, 16, 960) 24000 swish_60[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_61 (BatchNo (None, 16, 16, 960) 3840 depthwise_conv2d_21[0][0] \n__________________________________________________________________________________________________\nswish_61 (Swish) (None, 16, 16, 960) 0 batch_normalization_61[0][0] \n__________________________________________________________________________________________________\nlambda_21 (Lambda) (None, 1, 1, 960) 0 swish_61[0][0] \n__________________________________________________________________________________________________\nconv2d_81 (Conv2D) (None, 1, 1, 40) 38440 lambda_21[0][0] \n__________________________________________________________________________________________________\nswish_62 (Swish) (None, 1, 1, 40) 0 conv2d_81[0][0] \n__________________________________________________________________________________________________\nconv2d_82 (Conv2D) (None, 1, 1, 960) 39360 swish_62[0][0] \n__________________________________________________________________________________________________\nactivation_21 (Activation) (None, 1, 1, 960) 0 conv2d_82[0][0] \n__________________________________________________________________________________________________\nmultiply_21 (Multiply) (None, 16, 16, 960) 0 activation_21[0][0] \n swish_61[0][0] \n__________________________________________________________________________________________________\nconv2d_83 (Conv2D) (None, 16, 16, 160) 153600 multiply_21[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_62 (BatchNo (None, 16, 16, 160) 640 conv2d_83[0][0] \n__________________________________________________________________________________________________\ndrop_connect_16 (DropConnect) (None, 16, 16, 160) 0 batch_normalization_62[0][0] \n__________________________________________________________________________________________________\nadd_16 (Add) (None, 16, 16, 160) 0 drop_connect_16[0][0] \n add_15[0][0] \n__________________________________________________________________________________________________\nconv2d_84 (Conv2D) (None, 16, 16, 960) 153600 add_16[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_63 (BatchNo (None, 16, 16, 960) 3840 conv2d_84[0][0] \n__________________________________________________________________________________________________\nswish_63 (Swish) (None, 16, 16, 960) 0 batch_normalization_63[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_22 (DepthwiseC (None, 16, 16, 960) 24000 swish_63[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_64 (BatchNo (None, 16, 16, 960) 3840 depthwise_conv2d_22[0][0] \n__________________________________________________________________________________________________\nswish_64 (Swish) (None, 16, 16, 960) 0 batch_normalization_64[0][0] \n__________________________________________________________________________________________________\nlambda_22 (Lambda) (None, 1, 1, 960) 0 swish_64[0][0] \n__________________________________________________________________________________________________\nconv2d_85 (Conv2D) (None, 1, 1, 40) 38440 lambda_22[0][0] \n__________________________________________________________________________________________________\nswish_65 (Swish) (None, 1, 1, 40) 0 conv2d_85[0][0] \n__________________________________________________________________________________________________\nconv2d_86 (Conv2D) (None, 1, 1, 960) 39360 swish_65[0][0] \n__________________________________________________________________________________________________\nactivation_22 (Activation) (None, 1, 1, 960) 0 conv2d_86[0][0] \n__________________________________________________________________________________________________\nmultiply_22 (Multiply) (None, 16, 16, 960) 0 activation_22[0][0] \n swish_64[0][0] \n__________________________________________________________________________________________________\nconv2d_87 (Conv2D) (None, 16, 16, 160) 153600 multiply_22[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_65 (BatchNo (None, 16, 16, 160) 640 conv2d_87[0][0] \n__________________________________________________________________________________________________\ndrop_connect_17 (DropConnect) (None, 16, 16, 160) 0 batch_normalization_65[0][0] \n__________________________________________________________________________________________________\nadd_17 (Add) (None, 16, 16, 160) 0 drop_connect_17[0][0] \n add_16[0][0] \n__________________________________________________________________________________________________\nconv2d_88 (Conv2D) (None, 16, 16, 960) 153600 add_17[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_66 (BatchNo (None, 16, 16, 960) 3840 conv2d_88[0][0] \n__________________________________________________________________________________________________\nswish_66 (Swish) (None, 16, 16, 960) 0 batch_normalization_66[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_23 (DepthwiseC (None, 8, 8, 960) 24000 swish_66[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_67 (BatchNo (None, 8, 8, 960) 3840 depthwise_conv2d_23[0][0] \n__________________________________________________________________________________________________\nswish_67 (Swish) (None, 8, 8, 960) 0 batch_normalization_67[0][0] \n__________________________________________________________________________________________________\nlambda_23 (Lambda) (None, 1, 1, 960) 0 swish_67[0][0] \n__________________________________________________________________________________________________\nconv2d_89 (Conv2D) (None, 1, 1, 40) 38440 lambda_23[0][0] \n__________________________________________________________________________________________________\nswish_68 (Swish) (None, 1, 1, 40) 0 conv2d_89[0][0] \n__________________________________________________________________________________________________\nconv2d_90 (Conv2D) (None, 1, 1, 960) 39360 swish_68[0][0] \n__________________________________________________________________________________________________\nactivation_23 (Activation) (None, 1, 1, 960) 0 conv2d_90[0][0] \n__________________________________________________________________________________________________\nmultiply_23 (Multiply) (None, 8, 8, 960) 0 activation_23[0][0] \n swish_67[0][0] \n__________________________________________________________________________________________________\nconv2d_91 (Conv2D) (None, 8, 8, 272) 261120 multiply_23[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_68 (BatchNo (None, 8, 8, 272) 1088 conv2d_91[0][0] \n__________________________________________________________________________________________________\nconv2d_92 (Conv2D) (None, 8, 8, 1632) 443904 batch_normalization_68[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_69 (BatchNo (None, 8, 8, 1632) 6528 conv2d_92[0][0] \n__________________________________________________________________________________________________\nswish_69 (Swish) (None, 8, 8, 1632) 0 batch_normalization_69[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_24 (DepthwiseC (None, 8, 8, 1632) 40800 swish_69[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_70 (BatchNo (None, 8, 8, 1632) 6528 depthwise_conv2d_24[0][0] \n__________________________________________________________________________________________________\nswish_70 (Swish) (None, 8, 8, 1632) 0 batch_normalization_70[0][0] \n__________________________________________________________________________________________________\nlambda_24 (Lambda) (None, 1, 1, 1632) 0 swish_70[0][0] \n__________________________________________________________________________________________________\nconv2d_93 (Conv2D) (None, 1, 1, 68) 111044 lambda_24[0][0] \n__________________________________________________________________________________________________\nswish_71 (Swish) (None, 1, 1, 68) 0 conv2d_93[0][0] \n__________________________________________________________________________________________________\nconv2d_94 (Conv2D) (None, 1, 1, 1632) 112608 swish_71[0][0] \n__________________________________________________________________________________________________\nactivation_24 (Activation) (None, 1, 1, 1632) 0 conv2d_94[0][0] \n__________________________________________________________________________________________________\nmultiply_24 (Multiply) (None, 8, 8, 1632) 0 activation_24[0][0] \n swish_70[0][0] \n__________________________________________________________________________________________________\nconv2d_95 (Conv2D) (None, 8, 8, 272) 443904 multiply_24[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_71 (BatchNo (None, 8, 8, 272) 1088 conv2d_95[0][0] \n__________________________________________________________________________________________________\ndrop_connect_18 (DropConnect) (None, 8, 8, 272) 0 batch_normalization_71[0][0] \n__________________________________________________________________________________________________\nadd_18 (Add) (None, 8, 8, 272) 0 drop_connect_18[0][0] \n batch_normalization_68[0][0] \n__________________________________________________________________________________________________\nconv2d_96 (Conv2D) (None, 8, 8, 1632) 443904 add_18[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_72 (BatchNo (None, 8, 8, 1632) 6528 conv2d_96[0][0] \n__________________________________________________________________________________________________\nswish_72 (Swish) (None, 8, 8, 1632) 0 batch_normalization_72[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_25 (DepthwiseC (None, 8, 8, 1632) 40800 swish_72[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_73 (BatchNo (None, 8, 8, 1632) 6528 depthwise_conv2d_25[0][0] \n__________________________________________________________________________________________________\nswish_73 (Swish) (None, 8, 8, 1632) 0 batch_normalization_73[0][0] \n__________________________________________________________________________________________________\nlambda_25 (Lambda) (None, 1, 1, 1632) 0 swish_73[0][0] \n__________________________________________________________________________________________________\nconv2d_97 (Conv2D) (None, 1, 1, 68) 111044 lambda_25[0][0] \n__________________________________________________________________________________________________\nswish_74 (Swish) (None, 1, 1, 68) 0 conv2d_97[0][0] \n__________________________________________________________________________________________________\nconv2d_98 (Conv2D) (None, 1, 1, 1632) 112608 swish_74[0][0] \n__________________________________________________________________________________________________\nactivation_25 (Activation) (None, 1, 1, 1632) 0 conv2d_98[0][0] \n__________________________________________________________________________________________________\nmultiply_25 (Multiply) (None, 8, 8, 1632) 0 activation_25[0][0] \n swish_73[0][0] \n__________________________________________________________________________________________________\nconv2d_99 (Conv2D) (None, 8, 8, 272) 443904 multiply_25[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_74 (BatchNo (None, 8, 8, 272) 1088 conv2d_99[0][0] \n__________________________________________________________________________________________________\ndrop_connect_19 (DropConnect) (None, 8, 8, 272) 0 batch_normalization_74[0][0] \n__________________________________________________________________________________________________\nadd_19 (Add) (None, 8, 8, 272) 0 drop_connect_19[0][0] \n add_18[0][0] \n__________________________________________________________________________________________________\nconv2d_100 (Conv2D) (None, 8, 8, 1632) 443904 add_19[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_75 (BatchNo (None, 8, 8, 1632) 6528 conv2d_100[0][0] \n__________________________________________________________________________________________________\nswish_75 (Swish) (None, 8, 8, 1632) 0 batch_normalization_75[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_26 (DepthwiseC (None, 8, 8, 1632) 40800 swish_75[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_76 (BatchNo (None, 8, 8, 1632) 6528 depthwise_conv2d_26[0][0] \n__________________________________________________________________________________________________\nswish_76 (Swish) (None, 8, 8, 1632) 0 batch_normalization_76[0][0] \n__________________________________________________________________________________________________\nlambda_26 (Lambda) (None, 1, 1, 1632) 0 swish_76[0][0] \n__________________________________________________________________________________________________\nconv2d_101 (Conv2D) (None, 1, 1, 68) 111044 lambda_26[0][0] \n__________________________________________________________________________________________________\nswish_77 (Swish) (None, 1, 1, 68) 0 conv2d_101[0][0] \n__________________________________________________________________________________________________\nconv2d_102 (Conv2D) (None, 1, 1, 1632) 112608 swish_77[0][0] \n__________________________________________________________________________________________________\nactivation_26 (Activation) (None, 1, 1, 1632) 0 conv2d_102[0][0] \n__________________________________________________________________________________________________\nmultiply_26 (Multiply) (None, 8, 8, 1632) 0 activation_26[0][0] \n swish_76[0][0] \n__________________________________________________________________________________________________\nconv2d_103 (Conv2D) (None, 8, 8, 272) 443904 multiply_26[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_77 (BatchNo (None, 8, 8, 272) 1088 conv2d_103[0][0] \n__________________________________________________________________________________________________\ndrop_connect_20 (DropConnect) (None, 8, 8, 272) 0 batch_normalization_77[0][0] \n__________________________________________________________________________________________________\nadd_20 (Add) (None, 8, 8, 272) 0 drop_connect_20[0][0] \n add_19[0][0] \n__________________________________________________________________________________________________\nconv2d_104 (Conv2D) (None, 8, 8, 1632) 443904 add_20[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_78 (BatchNo (None, 8, 8, 1632) 6528 conv2d_104[0][0] \n__________________________________________________________________________________________________\nswish_78 (Swish) (None, 8, 8, 1632) 0 batch_normalization_78[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_27 (DepthwiseC (None, 8, 8, 1632) 40800 swish_78[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_79 (BatchNo (None, 8, 8, 1632) 6528 depthwise_conv2d_27[0][0] \n__________________________________________________________________________________________________\nswish_79 (Swish) (None, 8, 8, 1632) 0 batch_normalization_79[0][0] \n__________________________________________________________________________________________________\nlambda_27 (Lambda) (None, 1, 1, 1632) 0 swish_79[0][0] \n__________________________________________________________________________________________________\nconv2d_105 (Conv2D) (None, 1, 1, 68) 111044 lambda_27[0][0] \n__________________________________________________________________________________________________\nswish_80 (Swish) (None, 1, 1, 68) 0 conv2d_105[0][0] \n__________________________________________________________________________________________________\nconv2d_106 (Conv2D) (None, 1, 1, 1632) 112608 swish_80[0][0] \n__________________________________________________________________________________________________\nactivation_27 (Activation) (None, 1, 1, 1632) 0 conv2d_106[0][0] \n__________________________________________________________________________________________________\nmultiply_27 (Multiply) (None, 8, 8, 1632) 0 activation_27[0][0] \n swish_79[0][0] \n__________________________________________________________________________________________________\nconv2d_107 (Conv2D) (None, 8, 8, 272) 443904 multiply_27[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_80 (BatchNo (None, 8, 8, 272) 1088 conv2d_107[0][0] \n__________________________________________________________________________________________________\ndrop_connect_21 (DropConnect) (None, 8, 8, 272) 0 batch_normalization_80[0][0] \n__________________________________________________________________________________________________\nadd_21 (Add) (None, 8, 8, 272) 0 drop_connect_21[0][0] \n add_20[0][0] \n__________________________________________________________________________________________________\nconv2d_108 (Conv2D) (None, 8, 8, 1632) 443904 add_21[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_81 (BatchNo (None, 8, 8, 1632) 6528 conv2d_108[0][0] \n__________________________________________________________________________________________________\nswish_81 (Swish) (None, 8, 8, 1632) 0 batch_normalization_81[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_28 (DepthwiseC (None, 8, 8, 1632) 40800 swish_81[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_82 (BatchNo (None, 8, 8, 1632) 6528 depthwise_conv2d_28[0][0] \n__________________________________________________________________________________________________\nswish_82 (Swish) (None, 8, 8, 1632) 0 batch_normalization_82[0][0] \n__________________________________________________________________________________________________\nlambda_28 (Lambda) (None, 1, 1, 1632) 0 swish_82[0][0] \n__________________________________________________________________________________________________\nconv2d_109 (Conv2D) (None, 1, 1, 68) 111044 lambda_28[0][0] \n__________________________________________________________________________________________________\nswish_83 (Swish) (None, 1, 1, 68) 0 conv2d_109[0][0] \n__________________________________________________________________________________________________\nconv2d_110 (Conv2D) (None, 1, 1, 1632) 112608 swish_83[0][0] \n__________________________________________________________________________________________________\nactivation_28 (Activation) (None, 1, 1, 1632) 0 conv2d_110[0][0] \n__________________________________________________________________________________________________\nmultiply_28 (Multiply) (None, 8, 8, 1632) 0 activation_28[0][0] \n swish_82[0][0] \n__________________________________________________________________________________________________\nconv2d_111 (Conv2D) (None, 8, 8, 272) 443904 multiply_28[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_83 (BatchNo (None, 8, 8, 272) 1088 conv2d_111[0][0] \n__________________________________________________________________________________________________\ndrop_connect_22 (DropConnect) (None, 8, 8, 272) 0 batch_normalization_83[0][0] \n__________________________________________________________________________________________________\nadd_22 (Add) (None, 8, 8, 272) 0 drop_connect_22[0][0] \n add_21[0][0] \n__________________________________________________________________________________________________\nconv2d_112 (Conv2D) (None, 8, 8, 1632) 443904 add_22[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_84 (BatchNo (None, 8, 8, 1632) 6528 conv2d_112[0][0] \n__________________________________________________________________________________________________\nswish_84 (Swish) (None, 8, 8, 1632) 0 batch_normalization_84[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_29 (DepthwiseC (None, 8, 8, 1632) 40800 swish_84[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_85 (BatchNo (None, 8, 8, 1632) 6528 depthwise_conv2d_29[0][0] \n__________________________________________________________________________________________________\nswish_85 (Swish) (None, 8, 8, 1632) 0 batch_normalization_85[0][0] \n__________________________________________________________________________________________________\nlambda_29 (Lambda) (None, 1, 1, 1632) 0 swish_85[0][0] \n__________________________________________________________________________________________________\nconv2d_113 (Conv2D) (None, 1, 1, 68) 111044 lambda_29[0][0] \n__________________________________________________________________________________________________\nswish_86 (Swish) (None, 1, 1, 68) 0 conv2d_113[0][0] \n__________________________________________________________________________________________________\nconv2d_114 (Conv2D) (None, 1, 1, 1632) 112608 swish_86[0][0] \n__________________________________________________________________________________________________\nactivation_29 (Activation) (None, 1, 1, 1632) 0 conv2d_114[0][0] \n__________________________________________________________________________________________________\nmultiply_29 (Multiply) (None, 8, 8, 1632) 0 activation_29[0][0] \n swish_85[0][0] \n__________________________________________________________________________________________________\nconv2d_115 (Conv2D) (None, 8, 8, 272) 443904 multiply_29[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_86 (BatchNo (None, 8, 8, 272) 1088 conv2d_115[0][0] \n__________________________________________________________________________________________________\ndrop_connect_23 (DropConnect) (None, 8, 8, 272) 0 batch_normalization_86[0][0] \n__________________________________________________________________________________________________\nadd_23 (Add) (None, 8, 8, 272) 0 drop_connect_23[0][0] \n add_22[0][0] \n__________________________________________________________________________________________________\nconv2d_116 (Conv2D) (None, 8, 8, 1632) 443904 add_23[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_87 (BatchNo (None, 8, 8, 1632) 6528 conv2d_116[0][0] \n__________________________________________________________________________________________________\nswish_87 (Swish) (None, 8, 8, 1632) 0 batch_normalization_87[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_30 (DepthwiseC (None, 8, 8, 1632) 40800 swish_87[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_88 (BatchNo (None, 8, 8, 1632) 6528 depthwise_conv2d_30[0][0] \n__________________________________________________________________________________________________\nswish_88 (Swish) (None, 8, 8, 1632) 0 batch_normalization_88[0][0] \n__________________________________________________________________________________________________\nlambda_30 (Lambda) (None, 1, 1, 1632) 0 swish_88[0][0] \n__________________________________________________________________________________________________\nconv2d_117 (Conv2D) (None, 1, 1, 68) 111044 lambda_30[0][0] \n__________________________________________________________________________________________________\nswish_89 (Swish) (None, 1, 1, 68) 0 conv2d_117[0][0] \n__________________________________________________________________________________________________\nconv2d_118 (Conv2D) (None, 1, 1, 1632) 112608 swish_89[0][0] \n__________________________________________________________________________________________________\nactivation_30 (Activation) (None, 1, 1, 1632) 0 conv2d_118[0][0] \n__________________________________________________________________________________________________\nmultiply_30 (Multiply) (None, 8, 8, 1632) 0 activation_30[0][0] \n swish_88[0][0] \n__________________________________________________________________________________________________\nconv2d_119 (Conv2D) (None, 8, 8, 272) 443904 multiply_30[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_89 (BatchNo (None, 8, 8, 272) 1088 conv2d_119[0][0] \n__________________________________________________________________________________________________\ndrop_connect_24 (DropConnect) (None, 8, 8, 272) 0 batch_normalization_89[0][0] \n__________________________________________________________________________________________________\nadd_24 (Add) (None, 8, 8, 272) 0 drop_connect_24[0][0] \n add_23[0][0] \n__________________________________________________________________________________________________\nconv2d_120 (Conv2D) (None, 8, 8, 1632) 443904 add_24[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_90 (BatchNo (None, 8, 8, 1632) 6528 conv2d_120[0][0] \n__________________________________________________________________________________________________\nswish_90 (Swish) (None, 8, 8, 1632) 0 batch_normalization_90[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_31 (DepthwiseC (None, 8, 8, 1632) 14688 swish_90[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_91 (BatchNo (None, 8, 8, 1632) 6528 depthwise_conv2d_31[0][0] \n__________________________________________________________________________________________________\nswish_91 (Swish) (None, 8, 8, 1632) 0 batch_normalization_91[0][0] \n__________________________________________________________________________________________________\nlambda_31 (Lambda) (None, 1, 1, 1632) 0 swish_91[0][0] \n__________________________________________________________________________________________________\nconv2d_121 (Conv2D) (None, 1, 1, 68) 111044 lambda_31[0][0] \n__________________________________________________________________________________________________\nswish_92 (Swish) (None, 1, 1, 68) 0 conv2d_121[0][0] \n__________________________________________________________________________________________________\nconv2d_122 (Conv2D) (None, 1, 1, 1632) 112608 swish_92[0][0] \n__________________________________________________________________________________________________\nactivation_31 (Activation) (None, 1, 1, 1632) 0 conv2d_122[0][0] \n__________________________________________________________________________________________________\nmultiply_31 (Multiply) (None, 8, 8, 1632) 0 activation_31[0][0] \n swish_91[0][0] \n__________________________________________________________________________________________________\nconv2d_123 (Conv2D) (None, 8, 8, 448) 731136 multiply_31[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_92 (BatchNo (None, 8, 8, 448) 1792 conv2d_123[0][0] \n__________________________________________________________________________________________________\nconv2d_124 (Conv2D) (None, 8, 8, 2688) 1204224 batch_normalization_92[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_93 (BatchNo (None, 8, 8, 2688) 10752 conv2d_124[0][0] \n__________________________________________________________________________________________________\nswish_93 (Swish) (None, 8, 8, 2688) 0 batch_normalization_93[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_32 (DepthwiseC (None, 8, 8, 2688) 24192 swish_93[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_94 (BatchNo (None, 8, 8, 2688) 10752 depthwise_conv2d_32[0][0] \n__________________________________________________________________________________________________\nswish_94 (Swish) (None, 8, 8, 2688) 0 batch_normalization_94[0][0] \n__________________________________________________________________________________________________\nlambda_32 (Lambda) (None, 1, 1, 2688) 0 swish_94[0][0] \n__________________________________________________________________________________________________\nconv2d_125 (Conv2D) (None, 1, 1, 112) 301168 lambda_32[0][0] \n__________________________________________________________________________________________________\nswish_95 (Swish) (None, 1, 1, 112) 0 conv2d_125[0][0] \n__________________________________________________________________________________________________\nconv2d_126 (Conv2D) (None, 1, 1, 2688) 303744 swish_95[0][0] \n__________________________________________________________________________________________________\nactivation_32 (Activation) (None, 1, 1, 2688) 0 conv2d_126[0][0] \n__________________________________________________________________________________________________\nmultiply_32 (Multiply) (None, 8, 8, 2688) 0 activation_32[0][0] \n swish_94[0][0] \n__________________________________________________________________________________________________\nconv2d_127 (Conv2D) (None, 8, 8, 448) 1204224 multiply_32[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_95 (BatchNo (None, 8, 8, 448) 1792 conv2d_127[0][0] \n__________________________________________________________________________________________________\ndrop_connect_25 (DropConnect) (None, 8, 8, 448) 0 batch_normalization_95[0][0] \n__________________________________________________________________________________________________\nadd_25 (Add) (None, 8, 8, 448) 0 drop_connect_25[0][0] \n batch_normalization_92[0][0] \n__________________________________________________________________________________________________\nconv2d_128 (Conv2D) (None, 8, 8, 1792) 802816 add_25[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_96 (BatchNo (None, 8, 8, 1792) 7168 conv2d_128[0][0] \n__________________________________________________________________________________________________\nswish_96 (Swish) (None, 8, 8, 1792) 0 batch_normalization_96[0][0] \n__________________________________________________________________________________________________\nglobal_average_pooling2d_1 (Glo (None, 1792) 0 swish_96[0][0] \n__________________________________________________________________________________________________\nfinal_output (Dense) (None, 1) 1793 global_average_pooling2d_1[0][0] \n==================================================================================================\nTotal params: 17,675,609\nTrainable params: 1,793\nNon-trainable params: 17,673,816\n__________________________________________________________________________________________________\n" ], [ "STEP_SIZE_TRAIN = train_generator.n//train_generator.batch_size\nSTEP_SIZE_VALID = valid_generator.n//valid_generator.batch_size\n\nhistory_warmup = model.fit_generator(generator=train_generator,\n steps_per_epoch=STEP_SIZE_TRAIN,\n validation_data=valid_generator,\n validation_steps=STEP_SIZE_VALID,\n epochs=WARMUP_EPOCHS,\n verbose=2).history", "Epoch 1/5\n - 282s - loss: 1.2568 - acc: 0.3093 - val_loss: 1.4649 - val_acc: 0.4148\nEpoch 2/5\n - 269s - loss: 1.0960 - acc: 0.3228 - val_loss: 1.2900 - val_acc: 0.2810\nEpoch 3/5\n - 267s - loss: 1.0743 - acc: 0.3280 - val_loss: 1.3244 - val_acc: 0.2967\nEpoch 4/5\n - 266s - loss: 1.0808 - acc: 0.3231 - val_loss: 1.2172 - val_acc: 0.3338\nEpoch 5/5\n - 267s - loss: 1.0612 - acc: 0.3245 - val_loss: 1.1476 - val_acc: 0.2853\n" ] ], [ [ "# Fine-tune the model", "_____no_output_____" ] ], [ [ "for layer in model.layers:\n layer.trainable = True\n \ncheckpoint = ModelCheckpoint(model_path, monitor='val_loss', mode='min', save_best_only=True, save_weights_only=True)\nes = EarlyStopping(monitor='val_loss', mode='min', patience=ES_PATIENCE, restore_best_weights=True, verbose=1)\ncosine_lr = WarmUpCosineDecayScheduler(learning_rate_base=LEARNING_RATE,\n total_steps=TOTAL_STEPS,\n warmup_learning_rate=0.0,\n warmup_steps=WARMUP_STEPS,\n hold_base_rate_steps=(3 * STEP_SIZE))\n\ncallback_list = [checkpoint, es, cosine_lr]\noptimizer = optimizers.Adam(lr=LEARNING_RATE)\nmodel.compile(optimizer=optimizer, loss='mean_squared_error', metrics=metric_list)\nmodel.summary()", "__________________________________________________________________________________________________\nLayer (type) Output Shape Param # Connected to \n==================================================================================================\ninput_1 (InputLayer) (None, 256, 256, 3) 0 \n__________________________________________________________________________________________________\nconv2d_1 (Conv2D) (None, 128, 128, 48) 1296 input_1[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_1 (BatchNor (None, 128, 128, 48) 192 conv2d_1[0][0] \n__________________________________________________________________________________________________\nswish_1 (Swish) (None, 128, 128, 48) 0 batch_normalization_1[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_1 (DepthwiseCo (None, 128, 128, 48) 432 swish_1[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_2 (BatchNor (None, 128, 128, 48) 192 depthwise_conv2d_1[0][0] \n__________________________________________________________________________________________________\nswish_2 (Swish) (None, 128, 128, 48) 0 batch_normalization_2[0][0] \n__________________________________________________________________________________________________\nlambda_1 (Lambda) (None, 1, 1, 48) 0 swish_2[0][0] \n__________________________________________________________________________________________________\nconv2d_2 (Conv2D) (None, 1, 1, 12) 588 lambda_1[0][0] \n__________________________________________________________________________________________________\nswish_3 (Swish) (None, 1, 1, 12) 0 conv2d_2[0][0] \n__________________________________________________________________________________________________\nconv2d_3 (Conv2D) (None, 1, 1, 48) 624 swish_3[0][0] \n__________________________________________________________________________________________________\nactivation_1 (Activation) (None, 1, 1, 48) 0 conv2d_3[0][0] \n__________________________________________________________________________________________________\nmultiply_1 (Multiply) (None, 128, 128, 48) 0 activation_1[0][0] \n swish_2[0][0] \n__________________________________________________________________________________________________\nconv2d_4 (Conv2D) (None, 128, 128, 24) 1152 multiply_1[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_3 (BatchNor (None, 128, 128, 24) 96 conv2d_4[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_2 (DepthwiseCo (None, 128, 128, 24) 216 batch_normalization_3[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_4 (BatchNor (None, 128, 128, 24) 96 depthwise_conv2d_2[0][0] \n__________________________________________________________________________________________________\nswish_4 (Swish) (None, 128, 128, 24) 0 batch_normalization_4[0][0] \n__________________________________________________________________________________________________\nlambda_2 (Lambda) (None, 1, 1, 24) 0 swish_4[0][0] \n__________________________________________________________________________________________________\nconv2d_5 (Conv2D) (None, 1, 1, 6) 150 lambda_2[0][0] \n__________________________________________________________________________________________________\nswish_5 (Swish) (None, 1, 1, 6) 0 conv2d_5[0][0] \n__________________________________________________________________________________________________\nconv2d_6 (Conv2D) (None, 1, 1, 24) 168 swish_5[0][0] \n__________________________________________________________________________________________________\nactivation_2 (Activation) (None, 1, 1, 24) 0 conv2d_6[0][0] \n__________________________________________________________________________________________________\nmultiply_2 (Multiply) (None, 128, 128, 24) 0 activation_2[0][0] \n swish_4[0][0] \n__________________________________________________________________________________________________\nconv2d_7 (Conv2D) (None, 128, 128, 24) 576 multiply_2[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_5 (BatchNor (None, 128, 128, 24) 96 conv2d_7[0][0] \n__________________________________________________________________________________________________\ndrop_connect_1 (DropConnect) (None, 128, 128, 24) 0 batch_normalization_5[0][0] \n__________________________________________________________________________________________________\nadd_1 (Add) (None, 128, 128, 24) 0 drop_connect_1[0][0] \n batch_normalization_3[0][0] \n__________________________________________________________________________________________________\nconv2d_8 (Conv2D) (None, 128, 128, 144 3456 add_1[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_6 (BatchNor (None, 128, 128, 144 576 conv2d_8[0][0] \n__________________________________________________________________________________________________\nswish_6 (Swish) (None, 128, 128, 144 0 batch_normalization_6[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_3 (DepthwiseCo (None, 64, 64, 144) 1296 swish_6[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_7 (BatchNor (None, 64, 64, 144) 576 depthwise_conv2d_3[0][0] \n__________________________________________________________________________________________________\nswish_7 (Swish) (None, 64, 64, 144) 0 batch_normalization_7[0][0] \n__________________________________________________________________________________________________\nlambda_3 (Lambda) (None, 1, 1, 144) 0 swish_7[0][0] \n__________________________________________________________________________________________________\nconv2d_9 (Conv2D) (None, 1, 1, 6) 870 lambda_3[0][0] \n__________________________________________________________________________________________________\nswish_8 (Swish) (None, 1, 1, 6) 0 conv2d_9[0][0] \n__________________________________________________________________________________________________\nconv2d_10 (Conv2D) (None, 1, 1, 144) 1008 swish_8[0][0] \n__________________________________________________________________________________________________\nactivation_3 (Activation) (None, 1, 1, 144) 0 conv2d_10[0][0] \n__________________________________________________________________________________________________\nmultiply_3 (Multiply) (None, 64, 64, 144) 0 activation_3[0][0] \n swish_7[0][0] \n__________________________________________________________________________________________________\nconv2d_11 (Conv2D) (None, 64, 64, 32) 4608 multiply_3[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_8 (BatchNor (None, 64, 64, 32) 128 conv2d_11[0][0] \n__________________________________________________________________________________________________\nconv2d_12 (Conv2D) (None, 64, 64, 192) 6144 batch_normalization_8[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_9 (BatchNor (None, 64, 64, 192) 768 conv2d_12[0][0] \n__________________________________________________________________________________________________\nswish_9 (Swish) (None, 64, 64, 192) 0 batch_normalization_9[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_4 (DepthwiseCo (None, 64, 64, 192) 1728 swish_9[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_10 (BatchNo (None, 64, 64, 192) 768 depthwise_conv2d_4[0][0] \n__________________________________________________________________________________________________\nswish_10 (Swish) (None, 64, 64, 192) 0 batch_normalization_10[0][0] \n__________________________________________________________________________________________________\nlambda_4 (Lambda) (None, 1, 1, 192) 0 swish_10[0][0] \n__________________________________________________________________________________________________\nconv2d_13 (Conv2D) (None, 1, 1, 8) 1544 lambda_4[0][0] \n__________________________________________________________________________________________________\nswish_11 (Swish) (None, 1, 1, 8) 0 conv2d_13[0][0] \n__________________________________________________________________________________________________\nconv2d_14 (Conv2D) (None, 1, 1, 192) 1728 swish_11[0][0] \n__________________________________________________________________________________________________\nactivation_4 (Activation) (None, 1, 1, 192) 0 conv2d_14[0][0] \n__________________________________________________________________________________________________\nmultiply_4 (Multiply) (None, 64, 64, 192) 0 activation_4[0][0] \n swish_10[0][0] \n__________________________________________________________________________________________________\nconv2d_15 (Conv2D) (None, 64, 64, 32) 6144 multiply_4[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_11 (BatchNo (None, 64, 64, 32) 128 conv2d_15[0][0] \n__________________________________________________________________________________________________\ndrop_connect_2 (DropConnect) (None, 64, 64, 32) 0 batch_normalization_11[0][0] \n__________________________________________________________________________________________________\nadd_2 (Add) (None, 64, 64, 32) 0 drop_connect_2[0][0] \n batch_normalization_8[0][0] \n__________________________________________________________________________________________________\nconv2d_16 (Conv2D) (None, 64, 64, 192) 6144 add_2[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_12 (BatchNo (None, 64, 64, 192) 768 conv2d_16[0][0] \n__________________________________________________________________________________________________\nswish_12 (Swish) (None, 64, 64, 192) 0 batch_normalization_12[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_5 (DepthwiseCo (None, 64, 64, 192) 1728 swish_12[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_13 (BatchNo (None, 64, 64, 192) 768 depthwise_conv2d_5[0][0] \n__________________________________________________________________________________________________\nswish_13 (Swish) (None, 64, 64, 192) 0 batch_normalization_13[0][0] \n__________________________________________________________________________________________________\nlambda_5 (Lambda) (None, 1, 1, 192) 0 swish_13[0][0] \n__________________________________________________________________________________________________\nconv2d_17 (Conv2D) (None, 1, 1, 8) 1544 lambda_5[0][0] \n__________________________________________________________________________________________________\nswish_14 (Swish) (None, 1, 1, 8) 0 conv2d_17[0][0] \n__________________________________________________________________________________________________\nconv2d_18 (Conv2D) (None, 1, 1, 192) 1728 swish_14[0][0] \n__________________________________________________________________________________________________\nactivation_5 (Activation) (None, 1, 1, 192) 0 conv2d_18[0][0] \n__________________________________________________________________________________________________\nmultiply_5 (Multiply) (None, 64, 64, 192) 0 activation_5[0][0] \n swish_13[0][0] \n__________________________________________________________________________________________________\nconv2d_19 (Conv2D) (None, 64, 64, 32) 6144 multiply_5[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_14 (BatchNo (None, 64, 64, 32) 128 conv2d_19[0][0] \n__________________________________________________________________________________________________\ndrop_connect_3 (DropConnect) (None, 64, 64, 32) 0 batch_normalization_14[0][0] \n__________________________________________________________________________________________________\nadd_3 (Add) (None, 64, 64, 32) 0 drop_connect_3[0][0] \n add_2[0][0] \n__________________________________________________________________________________________________\nconv2d_20 (Conv2D) (None, 64, 64, 192) 6144 add_3[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_15 (BatchNo (None, 64, 64, 192) 768 conv2d_20[0][0] \n__________________________________________________________________________________________________\nswish_15 (Swish) (None, 64, 64, 192) 0 batch_normalization_15[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_6 (DepthwiseCo (None, 64, 64, 192) 1728 swish_15[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_16 (BatchNo (None, 64, 64, 192) 768 depthwise_conv2d_6[0][0] \n__________________________________________________________________________________________________\nswish_16 (Swish) (None, 64, 64, 192) 0 batch_normalization_16[0][0] \n__________________________________________________________________________________________________\nlambda_6 (Lambda) (None, 1, 1, 192) 0 swish_16[0][0] \n__________________________________________________________________________________________________\nconv2d_21 (Conv2D) (None, 1, 1, 8) 1544 lambda_6[0][0] \n__________________________________________________________________________________________________\nswish_17 (Swish) (None, 1, 1, 8) 0 conv2d_21[0][0] \n__________________________________________________________________________________________________\nconv2d_22 (Conv2D) (None, 1, 1, 192) 1728 swish_17[0][0] \n__________________________________________________________________________________________________\nactivation_6 (Activation) (None, 1, 1, 192) 0 conv2d_22[0][0] \n__________________________________________________________________________________________________\nmultiply_6 (Multiply) (None, 64, 64, 192) 0 activation_6[0][0] \n swish_16[0][0] \n__________________________________________________________________________________________________\nconv2d_23 (Conv2D) (None, 64, 64, 32) 6144 multiply_6[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_17 (BatchNo (None, 64, 64, 32) 128 conv2d_23[0][0] \n__________________________________________________________________________________________________\ndrop_connect_4 (DropConnect) (None, 64, 64, 32) 0 batch_normalization_17[0][0] \n__________________________________________________________________________________________________\nadd_4 (Add) (None, 64, 64, 32) 0 drop_connect_4[0][0] \n add_3[0][0] \n__________________________________________________________________________________________________\nconv2d_24 (Conv2D) (None, 64, 64, 192) 6144 add_4[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_18 (BatchNo (None, 64, 64, 192) 768 conv2d_24[0][0] \n__________________________________________________________________________________________________\nswish_18 (Swish) (None, 64, 64, 192) 0 batch_normalization_18[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_7 (DepthwiseCo (None, 32, 32, 192) 4800 swish_18[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_19 (BatchNo (None, 32, 32, 192) 768 depthwise_conv2d_7[0][0] \n__________________________________________________________________________________________________\nswish_19 (Swish) (None, 32, 32, 192) 0 batch_normalization_19[0][0] \n__________________________________________________________________________________________________\nlambda_7 (Lambda) (None, 1, 1, 192) 0 swish_19[0][0] \n__________________________________________________________________________________________________\nconv2d_25 (Conv2D) (None, 1, 1, 8) 1544 lambda_7[0][0] \n__________________________________________________________________________________________________\nswish_20 (Swish) (None, 1, 1, 8) 0 conv2d_25[0][0] \n__________________________________________________________________________________________________\nconv2d_26 (Conv2D) (None, 1, 1, 192) 1728 swish_20[0][0] \n__________________________________________________________________________________________________\nactivation_7 (Activation) (None, 1, 1, 192) 0 conv2d_26[0][0] \n__________________________________________________________________________________________________\nmultiply_7 (Multiply) (None, 32, 32, 192) 0 activation_7[0][0] \n swish_19[0][0] \n__________________________________________________________________________________________________\nconv2d_27 (Conv2D) (None, 32, 32, 56) 10752 multiply_7[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_20 (BatchNo (None, 32, 32, 56) 224 conv2d_27[0][0] \n__________________________________________________________________________________________________\nconv2d_28 (Conv2D) (None, 32, 32, 336) 18816 batch_normalization_20[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_21 (BatchNo (None, 32, 32, 336) 1344 conv2d_28[0][0] \n__________________________________________________________________________________________________\nswish_21 (Swish) (None, 32, 32, 336) 0 batch_normalization_21[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_8 (DepthwiseCo (None, 32, 32, 336) 8400 swish_21[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_22 (BatchNo (None, 32, 32, 336) 1344 depthwise_conv2d_8[0][0] \n__________________________________________________________________________________________________\nswish_22 (Swish) (None, 32, 32, 336) 0 batch_normalization_22[0][0] \n__________________________________________________________________________________________________\nlambda_8 (Lambda) (None, 1, 1, 336) 0 swish_22[0][0] \n__________________________________________________________________________________________________\nconv2d_29 (Conv2D) (None, 1, 1, 14) 4718 lambda_8[0][0] \n__________________________________________________________________________________________________\nswish_23 (Swish) (None, 1, 1, 14) 0 conv2d_29[0][0] \n__________________________________________________________________________________________________\nconv2d_30 (Conv2D) (None, 1, 1, 336) 5040 swish_23[0][0] \n__________________________________________________________________________________________________\nactivation_8 (Activation) (None, 1, 1, 336) 0 conv2d_30[0][0] \n__________________________________________________________________________________________________\nmultiply_8 (Multiply) (None, 32, 32, 336) 0 activation_8[0][0] \n swish_22[0][0] \n__________________________________________________________________________________________________\nconv2d_31 (Conv2D) (None, 32, 32, 56) 18816 multiply_8[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_23 (BatchNo (None, 32, 32, 56) 224 conv2d_31[0][0] \n__________________________________________________________________________________________________\ndrop_connect_5 (DropConnect) (None, 32, 32, 56) 0 batch_normalization_23[0][0] \n__________________________________________________________________________________________________\nadd_5 (Add) (None, 32, 32, 56) 0 drop_connect_5[0][0] \n batch_normalization_20[0][0] \n__________________________________________________________________________________________________\nconv2d_32 (Conv2D) (None, 32, 32, 336) 18816 add_5[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_24 (BatchNo (None, 32, 32, 336) 1344 conv2d_32[0][0] \n__________________________________________________________________________________________________\nswish_24 (Swish) (None, 32, 32, 336) 0 batch_normalization_24[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_9 (DepthwiseCo (None, 32, 32, 336) 8400 swish_24[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_25 (BatchNo (None, 32, 32, 336) 1344 depthwise_conv2d_9[0][0] \n__________________________________________________________________________________________________\nswish_25 (Swish) (None, 32, 32, 336) 0 batch_normalization_25[0][0] \n__________________________________________________________________________________________________\nlambda_9 (Lambda) (None, 1, 1, 336) 0 swish_25[0][0] \n__________________________________________________________________________________________________\nconv2d_33 (Conv2D) (None, 1, 1, 14) 4718 lambda_9[0][0] \n__________________________________________________________________________________________________\nswish_26 (Swish) (None, 1, 1, 14) 0 conv2d_33[0][0] \n__________________________________________________________________________________________________\nconv2d_34 (Conv2D) (None, 1, 1, 336) 5040 swish_26[0][0] \n__________________________________________________________________________________________________\nactivation_9 (Activation) (None, 1, 1, 336) 0 conv2d_34[0][0] \n__________________________________________________________________________________________________\nmultiply_9 (Multiply) (None, 32, 32, 336) 0 activation_9[0][0] \n swish_25[0][0] \n__________________________________________________________________________________________________\nconv2d_35 (Conv2D) (None, 32, 32, 56) 18816 multiply_9[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_26 (BatchNo (None, 32, 32, 56) 224 conv2d_35[0][0] \n__________________________________________________________________________________________________\ndrop_connect_6 (DropConnect) (None, 32, 32, 56) 0 batch_normalization_26[0][0] \n__________________________________________________________________________________________________\nadd_6 (Add) (None, 32, 32, 56) 0 drop_connect_6[0][0] \n add_5[0][0] \n__________________________________________________________________________________________________\nconv2d_36 (Conv2D) (None, 32, 32, 336) 18816 add_6[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_27 (BatchNo (None, 32, 32, 336) 1344 conv2d_36[0][0] \n__________________________________________________________________________________________________\nswish_27 (Swish) (None, 32, 32, 336) 0 batch_normalization_27[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_10 (DepthwiseC (None, 32, 32, 336) 8400 swish_27[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_28 (BatchNo (None, 32, 32, 336) 1344 depthwise_conv2d_10[0][0] \n__________________________________________________________________________________________________\nswish_28 (Swish) (None, 32, 32, 336) 0 batch_normalization_28[0][0] \n__________________________________________________________________________________________________\nlambda_10 (Lambda) (None, 1, 1, 336) 0 swish_28[0][0] \n__________________________________________________________________________________________________\nconv2d_37 (Conv2D) (None, 1, 1, 14) 4718 lambda_10[0][0] \n__________________________________________________________________________________________________\nswish_29 (Swish) (None, 1, 1, 14) 0 conv2d_37[0][0] \n__________________________________________________________________________________________________\nconv2d_38 (Conv2D) (None, 1, 1, 336) 5040 swish_29[0][0] \n__________________________________________________________________________________________________\nactivation_10 (Activation) (None, 1, 1, 336) 0 conv2d_38[0][0] \n__________________________________________________________________________________________________\nmultiply_10 (Multiply) (None, 32, 32, 336) 0 activation_10[0][0] \n swish_28[0][0] \n__________________________________________________________________________________________________\nconv2d_39 (Conv2D) (None, 32, 32, 56) 18816 multiply_10[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_29 (BatchNo (None, 32, 32, 56) 224 conv2d_39[0][0] \n__________________________________________________________________________________________________\ndrop_connect_7 (DropConnect) (None, 32, 32, 56) 0 batch_normalization_29[0][0] \n__________________________________________________________________________________________________\nadd_7 (Add) (None, 32, 32, 56) 0 drop_connect_7[0][0] \n add_6[0][0] \n__________________________________________________________________________________________________\nconv2d_40 (Conv2D) (None, 32, 32, 336) 18816 add_7[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_30 (BatchNo (None, 32, 32, 336) 1344 conv2d_40[0][0] \n__________________________________________________________________________________________________\nswish_30 (Swish) (None, 32, 32, 336) 0 batch_normalization_30[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_11 (DepthwiseC (None, 16, 16, 336) 3024 swish_30[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_31 (BatchNo (None, 16, 16, 336) 1344 depthwise_conv2d_11[0][0] \n__________________________________________________________________________________________________\nswish_31 (Swish) (None, 16, 16, 336) 0 batch_normalization_31[0][0] \n__________________________________________________________________________________________________\nlambda_11 (Lambda) (None, 1, 1, 336) 0 swish_31[0][0] \n__________________________________________________________________________________________________\nconv2d_41 (Conv2D) (None, 1, 1, 14) 4718 lambda_11[0][0] \n__________________________________________________________________________________________________\nswish_32 (Swish) (None, 1, 1, 14) 0 conv2d_41[0][0] \n__________________________________________________________________________________________________\nconv2d_42 (Conv2D) (None, 1, 1, 336) 5040 swish_32[0][0] \n__________________________________________________________________________________________________\nactivation_11 (Activation) (None, 1, 1, 336) 0 conv2d_42[0][0] \n__________________________________________________________________________________________________\nmultiply_11 (Multiply) (None, 16, 16, 336) 0 activation_11[0][0] \n swish_31[0][0] \n__________________________________________________________________________________________________\nconv2d_43 (Conv2D) (None, 16, 16, 112) 37632 multiply_11[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_32 (BatchNo (None, 16, 16, 112) 448 conv2d_43[0][0] \n__________________________________________________________________________________________________\nconv2d_44 (Conv2D) (None, 16, 16, 672) 75264 batch_normalization_32[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_33 (BatchNo (None, 16, 16, 672) 2688 conv2d_44[0][0] \n__________________________________________________________________________________________________\nswish_33 (Swish) (None, 16, 16, 672) 0 batch_normalization_33[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_12 (DepthwiseC (None, 16, 16, 672) 6048 swish_33[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_34 (BatchNo (None, 16, 16, 672) 2688 depthwise_conv2d_12[0][0] \n__________________________________________________________________________________________________\nswish_34 (Swish) (None, 16, 16, 672) 0 batch_normalization_34[0][0] \n__________________________________________________________________________________________________\nlambda_12 (Lambda) (None, 1, 1, 672) 0 swish_34[0][0] \n__________________________________________________________________________________________________\nconv2d_45 (Conv2D) (None, 1, 1, 28) 18844 lambda_12[0][0] \n__________________________________________________________________________________________________\nswish_35 (Swish) (None, 1, 1, 28) 0 conv2d_45[0][0] \n__________________________________________________________________________________________________\nconv2d_46 (Conv2D) (None, 1, 1, 672) 19488 swish_35[0][0] \n__________________________________________________________________________________________________\nactivation_12 (Activation) (None, 1, 1, 672) 0 conv2d_46[0][0] \n__________________________________________________________________________________________________\nmultiply_12 (Multiply) (None, 16, 16, 672) 0 activation_12[0][0] \n swish_34[0][0] \n__________________________________________________________________________________________________\nconv2d_47 (Conv2D) (None, 16, 16, 112) 75264 multiply_12[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_35 (BatchNo (None, 16, 16, 112) 448 conv2d_47[0][0] \n__________________________________________________________________________________________________\ndrop_connect_8 (DropConnect) (None, 16, 16, 112) 0 batch_normalization_35[0][0] \n__________________________________________________________________________________________________\nadd_8 (Add) (None, 16, 16, 112) 0 drop_connect_8[0][0] \n batch_normalization_32[0][0] \n__________________________________________________________________________________________________\nconv2d_48 (Conv2D) (None, 16, 16, 672) 75264 add_8[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_36 (BatchNo (None, 16, 16, 672) 2688 conv2d_48[0][0] \n__________________________________________________________________________________________________\nswish_36 (Swish) (None, 16, 16, 672) 0 batch_normalization_36[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_13 (DepthwiseC (None, 16, 16, 672) 6048 swish_36[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_37 (BatchNo (None, 16, 16, 672) 2688 depthwise_conv2d_13[0][0] \n__________________________________________________________________________________________________\nswish_37 (Swish) (None, 16, 16, 672) 0 batch_normalization_37[0][0] \n__________________________________________________________________________________________________\nlambda_13 (Lambda) (None, 1, 1, 672) 0 swish_37[0][0] \n__________________________________________________________________________________________________\nconv2d_49 (Conv2D) (None, 1, 1, 28) 18844 lambda_13[0][0] \n__________________________________________________________________________________________________\nswish_38 (Swish) (None, 1, 1, 28) 0 conv2d_49[0][0] \n__________________________________________________________________________________________________\nconv2d_50 (Conv2D) (None, 1, 1, 672) 19488 swish_38[0][0] \n__________________________________________________________________________________________________\nactivation_13 (Activation) (None, 1, 1, 672) 0 conv2d_50[0][0] \n__________________________________________________________________________________________________\nmultiply_13 (Multiply) (None, 16, 16, 672) 0 activation_13[0][0] \n swish_37[0][0] \n__________________________________________________________________________________________________\nconv2d_51 (Conv2D) (None, 16, 16, 112) 75264 multiply_13[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_38 (BatchNo (None, 16, 16, 112) 448 conv2d_51[0][0] \n__________________________________________________________________________________________________\ndrop_connect_9 (DropConnect) (None, 16, 16, 112) 0 batch_normalization_38[0][0] \n__________________________________________________________________________________________________\nadd_9 (Add) (None, 16, 16, 112) 0 drop_connect_9[0][0] \n add_8[0][0] \n__________________________________________________________________________________________________\nconv2d_52 (Conv2D) (None, 16, 16, 672) 75264 add_9[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_39 (BatchNo (None, 16, 16, 672) 2688 conv2d_52[0][0] \n__________________________________________________________________________________________________\nswish_39 (Swish) (None, 16, 16, 672) 0 batch_normalization_39[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_14 (DepthwiseC (None, 16, 16, 672) 6048 swish_39[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_40 (BatchNo (None, 16, 16, 672) 2688 depthwise_conv2d_14[0][0] \n__________________________________________________________________________________________________\nswish_40 (Swish) (None, 16, 16, 672) 0 batch_normalization_40[0][0] \n__________________________________________________________________________________________________\nlambda_14 (Lambda) (None, 1, 1, 672) 0 swish_40[0][0] \n__________________________________________________________________________________________________\nconv2d_53 (Conv2D) (None, 1, 1, 28) 18844 lambda_14[0][0] \n__________________________________________________________________________________________________\nswish_41 (Swish) (None, 1, 1, 28) 0 conv2d_53[0][0] \n__________________________________________________________________________________________________\nconv2d_54 (Conv2D) (None, 1, 1, 672) 19488 swish_41[0][0] \n__________________________________________________________________________________________________\nactivation_14 (Activation) (None, 1, 1, 672) 0 conv2d_54[0][0] \n__________________________________________________________________________________________________\nmultiply_14 (Multiply) (None, 16, 16, 672) 0 activation_14[0][0] \n swish_40[0][0] \n__________________________________________________________________________________________________\nconv2d_55 (Conv2D) (None, 16, 16, 112) 75264 multiply_14[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_41 (BatchNo (None, 16, 16, 112) 448 conv2d_55[0][0] \n__________________________________________________________________________________________________\ndrop_connect_10 (DropConnect) (None, 16, 16, 112) 0 batch_normalization_41[0][0] \n__________________________________________________________________________________________________\nadd_10 (Add) (None, 16, 16, 112) 0 drop_connect_10[0][0] \n add_9[0][0] \n__________________________________________________________________________________________________\nconv2d_56 (Conv2D) (None, 16, 16, 672) 75264 add_10[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_42 (BatchNo (None, 16, 16, 672) 2688 conv2d_56[0][0] \n__________________________________________________________________________________________________\nswish_42 (Swish) (None, 16, 16, 672) 0 batch_normalization_42[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_15 (DepthwiseC (None, 16, 16, 672) 6048 swish_42[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_43 (BatchNo (None, 16, 16, 672) 2688 depthwise_conv2d_15[0][0] \n__________________________________________________________________________________________________\nswish_43 (Swish) (None, 16, 16, 672) 0 batch_normalization_43[0][0] \n__________________________________________________________________________________________________\nlambda_15 (Lambda) (None, 1, 1, 672) 0 swish_43[0][0] \n__________________________________________________________________________________________________\nconv2d_57 (Conv2D) (None, 1, 1, 28) 18844 lambda_15[0][0] \n__________________________________________________________________________________________________\nswish_44 (Swish) (None, 1, 1, 28) 0 conv2d_57[0][0] \n__________________________________________________________________________________________________\nconv2d_58 (Conv2D) (None, 1, 1, 672) 19488 swish_44[0][0] \n__________________________________________________________________________________________________\nactivation_15 (Activation) (None, 1, 1, 672) 0 conv2d_58[0][0] \n__________________________________________________________________________________________________\nmultiply_15 (Multiply) (None, 16, 16, 672) 0 activation_15[0][0] \n swish_43[0][0] \n__________________________________________________________________________________________________\nconv2d_59 (Conv2D) (None, 16, 16, 112) 75264 multiply_15[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_44 (BatchNo (None, 16, 16, 112) 448 conv2d_59[0][0] \n__________________________________________________________________________________________________\ndrop_connect_11 (DropConnect) (None, 16, 16, 112) 0 batch_normalization_44[0][0] \n__________________________________________________________________________________________________\nadd_11 (Add) (None, 16, 16, 112) 0 drop_connect_11[0][0] \n add_10[0][0] \n__________________________________________________________________________________________________\nconv2d_60 (Conv2D) (None, 16, 16, 672) 75264 add_11[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_45 (BatchNo (None, 16, 16, 672) 2688 conv2d_60[0][0] \n__________________________________________________________________________________________________\nswish_45 (Swish) (None, 16, 16, 672) 0 batch_normalization_45[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_16 (DepthwiseC (None, 16, 16, 672) 6048 swish_45[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_46 (BatchNo (None, 16, 16, 672) 2688 depthwise_conv2d_16[0][0] \n__________________________________________________________________________________________________\nswish_46 (Swish) (None, 16, 16, 672) 0 batch_normalization_46[0][0] \n__________________________________________________________________________________________________\nlambda_16 (Lambda) (None, 1, 1, 672) 0 swish_46[0][0] \n__________________________________________________________________________________________________\nconv2d_61 (Conv2D) (None, 1, 1, 28) 18844 lambda_16[0][0] \n__________________________________________________________________________________________________\nswish_47 (Swish) (None, 1, 1, 28) 0 conv2d_61[0][0] \n__________________________________________________________________________________________________\nconv2d_62 (Conv2D) (None, 1, 1, 672) 19488 swish_47[0][0] \n__________________________________________________________________________________________________\nactivation_16 (Activation) (None, 1, 1, 672) 0 conv2d_62[0][0] \n__________________________________________________________________________________________________\nmultiply_16 (Multiply) (None, 16, 16, 672) 0 activation_16[0][0] \n swish_46[0][0] \n__________________________________________________________________________________________________\nconv2d_63 (Conv2D) (None, 16, 16, 112) 75264 multiply_16[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_47 (BatchNo (None, 16, 16, 112) 448 conv2d_63[0][0] \n__________________________________________________________________________________________________\ndrop_connect_12 (DropConnect) (None, 16, 16, 112) 0 batch_normalization_47[0][0] \n__________________________________________________________________________________________________\nadd_12 (Add) (None, 16, 16, 112) 0 drop_connect_12[0][0] \n add_11[0][0] \n__________________________________________________________________________________________________\nconv2d_64 (Conv2D) (None, 16, 16, 672) 75264 add_12[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_48 (BatchNo (None, 16, 16, 672) 2688 conv2d_64[0][0] \n__________________________________________________________________________________________________\nswish_48 (Swish) (None, 16, 16, 672) 0 batch_normalization_48[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_17 (DepthwiseC (None, 16, 16, 672) 16800 swish_48[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_49 (BatchNo (None, 16, 16, 672) 2688 depthwise_conv2d_17[0][0] \n__________________________________________________________________________________________________\nswish_49 (Swish) (None, 16, 16, 672) 0 batch_normalization_49[0][0] \n__________________________________________________________________________________________________\nlambda_17 (Lambda) (None, 1, 1, 672) 0 swish_49[0][0] \n__________________________________________________________________________________________________\nconv2d_65 (Conv2D) (None, 1, 1, 28) 18844 lambda_17[0][0] \n__________________________________________________________________________________________________\nswish_50 (Swish) (None, 1, 1, 28) 0 conv2d_65[0][0] \n__________________________________________________________________________________________________\nconv2d_66 (Conv2D) (None, 1, 1, 672) 19488 swish_50[0][0] \n__________________________________________________________________________________________________\nactivation_17 (Activation) (None, 1, 1, 672) 0 conv2d_66[0][0] \n__________________________________________________________________________________________________\nmultiply_17 (Multiply) (None, 16, 16, 672) 0 activation_17[0][0] \n swish_49[0][0] \n__________________________________________________________________________________________________\nconv2d_67 (Conv2D) (None, 16, 16, 160) 107520 multiply_17[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_50 (BatchNo (None, 16, 16, 160) 640 conv2d_67[0][0] \n__________________________________________________________________________________________________\nconv2d_68 (Conv2D) (None, 16, 16, 960) 153600 batch_normalization_50[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_51 (BatchNo (None, 16, 16, 960) 3840 conv2d_68[0][0] \n__________________________________________________________________________________________________\nswish_51 (Swish) (None, 16, 16, 960) 0 batch_normalization_51[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_18 (DepthwiseC (None, 16, 16, 960) 24000 swish_51[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_52 (BatchNo (None, 16, 16, 960) 3840 depthwise_conv2d_18[0][0] \n__________________________________________________________________________________________________\nswish_52 (Swish) (None, 16, 16, 960) 0 batch_normalization_52[0][0] \n__________________________________________________________________________________________________\nlambda_18 (Lambda) (None, 1, 1, 960) 0 swish_52[0][0] \n__________________________________________________________________________________________________\nconv2d_69 (Conv2D) (None, 1, 1, 40) 38440 lambda_18[0][0] \n__________________________________________________________________________________________________\nswish_53 (Swish) (None, 1, 1, 40) 0 conv2d_69[0][0] \n__________________________________________________________________________________________________\nconv2d_70 (Conv2D) (None, 1, 1, 960) 39360 swish_53[0][0] \n__________________________________________________________________________________________________\nactivation_18 (Activation) (None, 1, 1, 960) 0 conv2d_70[0][0] \n__________________________________________________________________________________________________\nmultiply_18 (Multiply) (None, 16, 16, 960) 0 activation_18[0][0] \n swish_52[0][0] \n__________________________________________________________________________________________________\nconv2d_71 (Conv2D) (None, 16, 16, 160) 153600 multiply_18[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_53 (BatchNo (None, 16, 16, 160) 640 conv2d_71[0][0] \n__________________________________________________________________________________________________\ndrop_connect_13 (DropConnect) (None, 16, 16, 160) 0 batch_normalization_53[0][0] \n__________________________________________________________________________________________________\nadd_13 (Add) (None, 16, 16, 160) 0 drop_connect_13[0][0] \n batch_normalization_50[0][0] \n__________________________________________________________________________________________________\nconv2d_72 (Conv2D) (None, 16, 16, 960) 153600 add_13[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_54 (BatchNo (None, 16, 16, 960) 3840 conv2d_72[0][0] \n__________________________________________________________________________________________________\nswish_54 (Swish) (None, 16, 16, 960) 0 batch_normalization_54[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_19 (DepthwiseC (None, 16, 16, 960) 24000 swish_54[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_55 (BatchNo (None, 16, 16, 960) 3840 depthwise_conv2d_19[0][0] \n__________________________________________________________________________________________________\nswish_55 (Swish) (None, 16, 16, 960) 0 batch_normalization_55[0][0] \n__________________________________________________________________________________________________\nlambda_19 (Lambda) (None, 1, 1, 960) 0 swish_55[0][0] \n__________________________________________________________________________________________________\nconv2d_73 (Conv2D) (None, 1, 1, 40) 38440 lambda_19[0][0] \n__________________________________________________________________________________________________\nswish_56 (Swish) (None, 1, 1, 40) 0 conv2d_73[0][0] \n__________________________________________________________________________________________________\nconv2d_74 (Conv2D) (None, 1, 1, 960) 39360 swish_56[0][0] \n__________________________________________________________________________________________________\nactivation_19 (Activation) (None, 1, 1, 960) 0 conv2d_74[0][0] \n__________________________________________________________________________________________________\nmultiply_19 (Multiply) (None, 16, 16, 960) 0 activation_19[0][0] \n swish_55[0][0] \n__________________________________________________________________________________________________\nconv2d_75 (Conv2D) (None, 16, 16, 160) 153600 multiply_19[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_56 (BatchNo (None, 16, 16, 160) 640 conv2d_75[0][0] \n__________________________________________________________________________________________________\ndrop_connect_14 (DropConnect) (None, 16, 16, 160) 0 batch_normalization_56[0][0] \n__________________________________________________________________________________________________\nadd_14 (Add) (None, 16, 16, 160) 0 drop_connect_14[0][0] \n add_13[0][0] \n__________________________________________________________________________________________________\nconv2d_76 (Conv2D) (None, 16, 16, 960) 153600 add_14[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_57 (BatchNo (None, 16, 16, 960) 3840 conv2d_76[0][0] \n__________________________________________________________________________________________________\nswish_57 (Swish) (None, 16, 16, 960) 0 batch_normalization_57[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_20 (DepthwiseC (None, 16, 16, 960) 24000 swish_57[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_58 (BatchNo (None, 16, 16, 960) 3840 depthwise_conv2d_20[0][0] \n__________________________________________________________________________________________________\nswish_58 (Swish) (None, 16, 16, 960) 0 batch_normalization_58[0][0] \n__________________________________________________________________________________________________\nlambda_20 (Lambda) (None, 1, 1, 960) 0 swish_58[0][0] \n__________________________________________________________________________________________________\nconv2d_77 (Conv2D) (None, 1, 1, 40) 38440 lambda_20[0][0] \n__________________________________________________________________________________________________\nswish_59 (Swish) (None, 1, 1, 40) 0 conv2d_77[0][0] \n__________________________________________________________________________________________________\nconv2d_78 (Conv2D) (None, 1, 1, 960) 39360 swish_59[0][0] \n__________________________________________________________________________________________________\nactivation_20 (Activation) (None, 1, 1, 960) 0 conv2d_78[0][0] \n__________________________________________________________________________________________________\nmultiply_20 (Multiply) (None, 16, 16, 960) 0 activation_20[0][0] \n swish_58[0][0] \n__________________________________________________________________________________________________\nconv2d_79 (Conv2D) (None, 16, 16, 160) 153600 multiply_20[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_59 (BatchNo (None, 16, 16, 160) 640 conv2d_79[0][0] \n__________________________________________________________________________________________________\ndrop_connect_15 (DropConnect) (None, 16, 16, 160) 0 batch_normalization_59[0][0] \n__________________________________________________________________________________________________\nadd_15 (Add) (None, 16, 16, 160) 0 drop_connect_15[0][0] \n add_14[0][0] \n__________________________________________________________________________________________________\nconv2d_80 (Conv2D) (None, 16, 16, 960) 153600 add_15[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_60 (BatchNo (None, 16, 16, 960) 3840 conv2d_80[0][0] \n__________________________________________________________________________________________________\nswish_60 (Swish) (None, 16, 16, 960) 0 batch_normalization_60[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_21 (DepthwiseC (None, 16, 16, 960) 24000 swish_60[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_61 (BatchNo (None, 16, 16, 960) 3840 depthwise_conv2d_21[0][0] \n__________________________________________________________________________________________________\nswish_61 (Swish) (None, 16, 16, 960) 0 batch_normalization_61[0][0] \n__________________________________________________________________________________________________\nlambda_21 (Lambda) (None, 1, 1, 960) 0 swish_61[0][0] \n__________________________________________________________________________________________________\nconv2d_81 (Conv2D) (None, 1, 1, 40) 38440 lambda_21[0][0] \n__________________________________________________________________________________________________\nswish_62 (Swish) (None, 1, 1, 40) 0 conv2d_81[0][0] \n__________________________________________________________________________________________________\nconv2d_82 (Conv2D) (None, 1, 1, 960) 39360 swish_62[0][0] \n__________________________________________________________________________________________________\nactivation_21 (Activation) (None, 1, 1, 960) 0 conv2d_82[0][0] \n__________________________________________________________________________________________________\nmultiply_21 (Multiply) (None, 16, 16, 960) 0 activation_21[0][0] \n swish_61[0][0] \n__________________________________________________________________________________________________\nconv2d_83 (Conv2D) (None, 16, 16, 160) 153600 multiply_21[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_62 (BatchNo (None, 16, 16, 160) 640 conv2d_83[0][0] \n__________________________________________________________________________________________________\ndrop_connect_16 (DropConnect) (None, 16, 16, 160) 0 batch_normalization_62[0][0] \n__________________________________________________________________________________________________\nadd_16 (Add) (None, 16, 16, 160) 0 drop_connect_16[0][0] \n add_15[0][0] \n__________________________________________________________________________________________________\nconv2d_84 (Conv2D) (None, 16, 16, 960) 153600 add_16[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_63 (BatchNo (None, 16, 16, 960) 3840 conv2d_84[0][0] \n__________________________________________________________________________________________________\nswish_63 (Swish) (None, 16, 16, 960) 0 batch_normalization_63[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_22 (DepthwiseC (None, 16, 16, 960) 24000 swish_63[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_64 (BatchNo (None, 16, 16, 960) 3840 depthwise_conv2d_22[0][0] \n__________________________________________________________________________________________________\nswish_64 (Swish) (None, 16, 16, 960) 0 batch_normalization_64[0][0] \n__________________________________________________________________________________________________\nlambda_22 (Lambda) (None, 1, 1, 960) 0 swish_64[0][0] \n__________________________________________________________________________________________________\nconv2d_85 (Conv2D) (None, 1, 1, 40) 38440 lambda_22[0][0] \n__________________________________________________________________________________________________\nswish_65 (Swish) (None, 1, 1, 40) 0 conv2d_85[0][0] \n__________________________________________________________________________________________________\nconv2d_86 (Conv2D) (None, 1, 1, 960) 39360 swish_65[0][0] \n__________________________________________________________________________________________________\nactivation_22 (Activation) (None, 1, 1, 960) 0 conv2d_86[0][0] \n__________________________________________________________________________________________________\nmultiply_22 (Multiply) (None, 16, 16, 960) 0 activation_22[0][0] \n swish_64[0][0] \n__________________________________________________________________________________________________\nconv2d_87 (Conv2D) (None, 16, 16, 160) 153600 multiply_22[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_65 (BatchNo (None, 16, 16, 160) 640 conv2d_87[0][0] \n__________________________________________________________________________________________________\ndrop_connect_17 (DropConnect) (None, 16, 16, 160) 0 batch_normalization_65[0][0] \n__________________________________________________________________________________________________\nadd_17 (Add) (None, 16, 16, 160) 0 drop_connect_17[0][0] \n add_16[0][0] \n__________________________________________________________________________________________________\nconv2d_88 (Conv2D) (None, 16, 16, 960) 153600 add_17[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_66 (BatchNo (None, 16, 16, 960) 3840 conv2d_88[0][0] \n__________________________________________________________________________________________________\nswish_66 (Swish) (None, 16, 16, 960) 0 batch_normalization_66[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_23 (DepthwiseC (None, 8, 8, 960) 24000 swish_66[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_67 (BatchNo (None, 8, 8, 960) 3840 depthwise_conv2d_23[0][0] \n__________________________________________________________________________________________________\nswish_67 (Swish) (None, 8, 8, 960) 0 batch_normalization_67[0][0] \n__________________________________________________________________________________________________\nlambda_23 (Lambda) (None, 1, 1, 960) 0 swish_67[0][0] \n__________________________________________________________________________________________________\nconv2d_89 (Conv2D) (None, 1, 1, 40) 38440 lambda_23[0][0] \n__________________________________________________________________________________________________\nswish_68 (Swish) (None, 1, 1, 40) 0 conv2d_89[0][0] \n__________________________________________________________________________________________________\nconv2d_90 (Conv2D) (None, 1, 1, 960) 39360 swish_68[0][0] \n__________________________________________________________________________________________________\nactivation_23 (Activation) (None, 1, 1, 960) 0 conv2d_90[0][0] \n__________________________________________________________________________________________________\nmultiply_23 (Multiply) (None, 8, 8, 960) 0 activation_23[0][0] \n swish_67[0][0] \n__________________________________________________________________________________________________\nconv2d_91 (Conv2D) (None, 8, 8, 272) 261120 multiply_23[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_68 (BatchNo (None, 8, 8, 272) 1088 conv2d_91[0][0] \n__________________________________________________________________________________________________\nconv2d_92 (Conv2D) (None, 8, 8, 1632) 443904 batch_normalization_68[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_69 (BatchNo (None, 8, 8, 1632) 6528 conv2d_92[0][0] \n__________________________________________________________________________________________________\nswish_69 (Swish) (None, 8, 8, 1632) 0 batch_normalization_69[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_24 (DepthwiseC (None, 8, 8, 1632) 40800 swish_69[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_70 (BatchNo (None, 8, 8, 1632) 6528 depthwise_conv2d_24[0][0] \n__________________________________________________________________________________________________\nswish_70 (Swish) (None, 8, 8, 1632) 0 batch_normalization_70[0][0] \n__________________________________________________________________________________________________\nlambda_24 (Lambda) (None, 1, 1, 1632) 0 swish_70[0][0] \n__________________________________________________________________________________________________\nconv2d_93 (Conv2D) (None, 1, 1, 68) 111044 lambda_24[0][0] \n__________________________________________________________________________________________________\nswish_71 (Swish) (None, 1, 1, 68) 0 conv2d_93[0][0] \n__________________________________________________________________________________________________\nconv2d_94 (Conv2D) (None, 1, 1, 1632) 112608 swish_71[0][0] \n__________________________________________________________________________________________________\nactivation_24 (Activation) (None, 1, 1, 1632) 0 conv2d_94[0][0] \n__________________________________________________________________________________________________\nmultiply_24 (Multiply) (None, 8, 8, 1632) 0 activation_24[0][0] \n swish_70[0][0] \n__________________________________________________________________________________________________\nconv2d_95 (Conv2D) (None, 8, 8, 272) 443904 multiply_24[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_71 (BatchNo (None, 8, 8, 272) 1088 conv2d_95[0][0] \n__________________________________________________________________________________________________\ndrop_connect_18 (DropConnect) (None, 8, 8, 272) 0 batch_normalization_71[0][0] \n__________________________________________________________________________________________________\nadd_18 (Add) (None, 8, 8, 272) 0 drop_connect_18[0][0] \n batch_normalization_68[0][0] \n__________________________________________________________________________________________________\nconv2d_96 (Conv2D) (None, 8, 8, 1632) 443904 add_18[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_72 (BatchNo (None, 8, 8, 1632) 6528 conv2d_96[0][0] \n__________________________________________________________________________________________________\nswish_72 (Swish) (None, 8, 8, 1632) 0 batch_normalization_72[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_25 (DepthwiseC (None, 8, 8, 1632) 40800 swish_72[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_73 (BatchNo (None, 8, 8, 1632) 6528 depthwise_conv2d_25[0][0] \n__________________________________________________________________________________________________\nswish_73 (Swish) (None, 8, 8, 1632) 0 batch_normalization_73[0][0] \n__________________________________________________________________________________________________\nlambda_25 (Lambda) (None, 1, 1, 1632) 0 swish_73[0][0] \n__________________________________________________________________________________________________\nconv2d_97 (Conv2D) (None, 1, 1, 68) 111044 lambda_25[0][0] \n__________________________________________________________________________________________________\nswish_74 (Swish) (None, 1, 1, 68) 0 conv2d_97[0][0] \n__________________________________________________________________________________________________\nconv2d_98 (Conv2D) (None, 1, 1, 1632) 112608 swish_74[0][0] \n__________________________________________________________________________________________________\nactivation_25 (Activation) (None, 1, 1, 1632) 0 conv2d_98[0][0] \n__________________________________________________________________________________________________\nmultiply_25 (Multiply) (None, 8, 8, 1632) 0 activation_25[0][0] \n swish_73[0][0] \n__________________________________________________________________________________________________\nconv2d_99 (Conv2D) (None, 8, 8, 272) 443904 multiply_25[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_74 (BatchNo (None, 8, 8, 272) 1088 conv2d_99[0][0] \n__________________________________________________________________________________________________\ndrop_connect_19 (DropConnect) (None, 8, 8, 272) 0 batch_normalization_74[0][0] \n__________________________________________________________________________________________________\nadd_19 (Add) (None, 8, 8, 272) 0 drop_connect_19[0][0] \n add_18[0][0] \n__________________________________________________________________________________________________\nconv2d_100 (Conv2D) (None, 8, 8, 1632) 443904 add_19[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_75 (BatchNo (None, 8, 8, 1632) 6528 conv2d_100[0][0] \n__________________________________________________________________________________________________\nswish_75 (Swish) (None, 8, 8, 1632) 0 batch_normalization_75[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_26 (DepthwiseC (None, 8, 8, 1632) 40800 swish_75[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_76 (BatchNo (None, 8, 8, 1632) 6528 depthwise_conv2d_26[0][0] \n__________________________________________________________________________________________________\nswish_76 (Swish) (None, 8, 8, 1632) 0 batch_normalization_76[0][0] \n__________________________________________________________________________________________________\nlambda_26 (Lambda) (None, 1, 1, 1632) 0 swish_76[0][0] \n__________________________________________________________________________________________________\nconv2d_101 (Conv2D) (None, 1, 1, 68) 111044 lambda_26[0][0] \n__________________________________________________________________________________________________\nswish_77 (Swish) (None, 1, 1, 68) 0 conv2d_101[0][0] \n__________________________________________________________________________________________________\nconv2d_102 (Conv2D) (None, 1, 1, 1632) 112608 swish_77[0][0] \n__________________________________________________________________________________________________\nactivation_26 (Activation) (None, 1, 1, 1632) 0 conv2d_102[0][0] \n__________________________________________________________________________________________________\nmultiply_26 (Multiply) (None, 8, 8, 1632) 0 activation_26[0][0] \n swish_76[0][0] \n__________________________________________________________________________________________________\nconv2d_103 (Conv2D) (None, 8, 8, 272) 443904 multiply_26[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_77 (BatchNo (None, 8, 8, 272) 1088 conv2d_103[0][0] \n__________________________________________________________________________________________________\ndrop_connect_20 (DropConnect) (None, 8, 8, 272) 0 batch_normalization_77[0][0] \n__________________________________________________________________________________________________\nadd_20 (Add) (None, 8, 8, 272) 0 drop_connect_20[0][0] \n add_19[0][0] \n__________________________________________________________________________________________________\nconv2d_104 (Conv2D) (None, 8, 8, 1632) 443904 add_20[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_78 (BatchNo (None, 8, 8, 1632) 6528 conv2d_104[0][0] \n__________________________________________________________________________________________________\nswish_78 (Swish) (None, 8, 8, 1632) 0 batch_normalization_78[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_27 (DepthwiseC (None, 8, 8, 1632) 40800 swish_78[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_79 (BatchNo (None, 8, 8, 1632) 6528 depthwise_conv2d_27[0][0] \n__________________________________________________________________________________________________\nswish_79 (Swish) (None, 8, 8, 1632) 0 batch_normalization_79[0][0] \n__________________________________________________________________________________________________\nlambda_27 (Lambda) (None, 1, 1, 1632) 0 swish_79[0][0] \n__________________________________________________________________________________________________\nconv2d_105 (Conv2D) (None, 1, 1, 68) 111044 lambda_27[0][0] \n__________________________________________________________________________________________________\nswish_80 (Swish) (None, 1, 1, 68) 0 conv2d_105[0][0] \n__________________________________________________________________________________________________\nconv2d_106 (Conv2D) (None, 1, 1, 1632) 112608 swish_80[0][0] \n__________________________________________________________________________________________________\nactivation_27 (Activation) (None, 1, 1, 1632) 0 conv2d_106[0][0] \n__________________________________________________________________________________________________\nmultiply_27 (Multiply) (None, 8, 8, 1632) 0 activation_27[0][0] \n swish_79[0][0] \n__________________________________________________________________________________________________\nconv2d_107 (Conv2D) (None, 8, 8, 272) 443904 multiply_27[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_80 (BatchNo (None, 8, 8, 272) 1088 conv2d_107[0][0] \n__________________________________________________________________________________________________\ndrop_connect_21 (DropConnect) (None, 8, 8, 272) 0 batch_normalization_80[0][0] \n__________________________________________________________________________________________________\nadd_21 (Add) (None, 8, 8, 272) 0 drop_connect_21[0][0] \n add_20[0][0] \n__________________________________________________________________________________________________\nconv2d_108 (Conv2D) (None, 8, 8, 1632) 443904 add_21[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_81 (BatchNo (None, 8, 8, 1632) 6528 conv2d_108[0][0] \n__________________________________________________________________________________________________\nswish_81 (Swish) (None, 8, 8, 1632) 0 batch_normalization_81[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_28 (DepthwiseC (None, 8, 8, 1632) 40800 swish_81[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_82 (BatchNo (None, 8, 8, 1632) 6528 depthwise_conv2d_28[0][0] \n__________________________________________________________________________________________________\nswish_82 (Swish) (None, 8, 8, 1632) 0 batch_normalization_82[0][0] \n__________________________________________________________________________________________________\nlambda_28 (Lambda) (None, 1, 1, 1632) 0 swish_82[0][0] \n__________________________________________________________________________________________________\nconv2d_109 (Conv2D) (None, 1, 1, 68) 111044 lambda_28[0][0] \n__________________________________________________________________________________________________\nswish_83 (Swish) (None, 1, 1, 68) 0 conv2d_109[0][0] \n__________________________________________________________________________________________________\nconv2d_110 (Conv2D) (None, 1, 1, 1632) 112608 swish_83[0][0] \n__________________________________________________________________________________________________\nactivation_28 (Activation) (None, 1, 1, 1632) 0 conv2d_110[0][0] \n__________________________________________________________________________________________________\nmultiply_28 (Multiply) (None, 8, 8, 1632) 0 activation_28[0][0] \n swish_82[0][0] \n__________________________________________________________________________________________________\nconv2d_111 (Conv2D) (None, 8, 8, 272) 443904 multiply_28[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_83 (BatchNo (None, 8, 8, 272) 1088 conv2d_111[0][0] \n__________________________________________________________________________________________________\ndrop_connect_22 (DropConnect) (None, 8, 8, 272) 0 batch_normalization_83[0][0] \n__________________________________________________________________________________________________\nadd_22 (Add) (None, 8, 8, 272) 0 drop_connect_22[0][0] \n add_21[0][0] \n__________________________________________________________________________________________________\nconv2d_112 (Conv2D) (None, 8, 8, 1632) 443904 add_22[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_84 (BatchNo (None, 8, 8, 1632) 6528 conv2d_112[0][0] \n__________________________________________________________________________________________________\nswish_84 (Swish) (None, 8, 8, 1632) 0 batch_normalization_84[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_29 (DepthwiseC (None, 8, 8, 1632) 40800 swish_84[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_85 (BatchNo (None, 8, 8, 1632) 6528 depthwise_conv2d_29[0][0] \n__________________________________________________________________________________________________\nswish_85 (Swish) (None, 8, 8, 1632) 0 batch_normalization_85[0][0] \n__________________________________________________________________________________________________\nlambda_29 (Lambda) (None, 1, 1, 1632) 0 swish_85[0][0] \n__________________________________________________________________________________________________\nconv2d_113 (Conv2D) (None, 1, 1, 68) 111044 lambda_29[0][0] \n__________________________________________________________________________________________________\nswish_86 (Swish) (None, 1, 1, 68) 0 conv2d_113[0][0] \n__________________________________________________________________________________________________\nconv2d_114 (Conv2D) (None, 1, 1, 1632) 112608 swish_86[0][0] \n__________________________________________________________________________________________________\nactivation_29 (Activation) (None, 1, 1, 1632) 0 conv2d_114[0][0] \n__________________________________________________________________________________________________\nmultiply_29 (Multiply) (None, 8, 8, 1632) 0 activation_29[0][0] \n swish_85[0][0] \n__________________________________________________________________________________________________\nconv2d_115 (Conv2D) (None, 8, 8, 272) 443904 multiply_29[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_86 (BatchNo (None, 8, 8, 272) 1088 conv2d_115[0][0] \n__________________________________________________________________________________________________\ndrop_connect_23 (DropConnect) (None, 8, 8, 272) 0 batch_normalization_86[0][0] \n__________________________________________________________________________________________________\nadd_23 (Add) (None, 8, 8, 272) 0 drop_connect_23[0][0] \n add_22[0][0] \n__________________________________________________________________________________________________\nconv2d_116 (Conv2D) (None, 8, 8, 1632) 443904 add_23[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_87 (BatchNo (None, 8, 8, 1632) 6528 conv2d_116[0][0] \n__________________________________________________________________________________________________\nswish_87 (Swish) (None, 8, 8, 1632) 0 batch_normalization_87[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_30 (DepthwiseC (None, 8, 8, 1632) 40800 swish_87[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_88 (BatchNo (None, 8, 8, 1632) 6528 depthwise_conv2d_30[0][0] \n__________________________________________________________________________________________________\nswish_88 (Swish) (None, 8, 8, 1632) 0 batch_normalization_88[0][0] \n__________________________________________________________________________________________________\nlambda_30 (Lambda) (None, 1, 1, 1632) 0 swish_88[0][0] \n__________________________________________________________________________________________________\nconv2d_117 (Conv2D) (None, 1, 1, 68) 111044 lambda_30[0][0] \n__________________________________________________________________________________________________\nswish_89 (Swish) (None, 1, 1, 68) 0 conv2d_117[0][0] \n__________________________________________________________________________________________________\nconv2d_118 (Conv2D) (None, 1, 1, 1632) 112608 swish_89[0][0] \n__________________________________________________________________________________________________\nactivation_30 (Activation) (None, 1, 1, 1632) 0 conv2d_118[0][0] \n__________________________________________________________________________________________________\nmultiply_30 (Multiply) (None, 8, 8, 1632) 0 activation_30[0][0] \n swish_88[0][0] \n__________________________________________________________________________________________________\nconv2d_119 (Conv2D) (None, 8, 8, 272) 443904 multiply_30[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_89 (BatchNo (None, 8, 8, 272) 1088 conv2d_119[0][0] \n__________________________________________________________________________________________________\ndrop_connect_24 (DropConnect) (None, 8, 8, 272) 0 batch_normalization_89[0][0] \n__________________________________________________________________________________________________\nadd_24 (Add) (None, 8, 8, 272) 0 drop_connect_24[0][0] \n add_23[0][0] \n__________________________________________________________________________________________________\nconv2d_120 (Conv2D) (None, 8, 8, 1632) 443904 add_24[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_90 (BatchNo (None, 8, 8, 1632) 6528 conv2d_120[0][0] \n__________________________________________________________________________________________________\nswish_90 (Swish) (None, 8, 8, 1632) 0 batch_normalization_90[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_31 (DepthwiseC (None, 8, 8, 1632) 14688 swish_90[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_91 (BatchNo (None, 8, 8, 1632) 6528 depthwise_conv2d_31[0][0] \n__________________________________________________________________________________________________\nswish_91 (Swish) (None, 8, 8, 1632) 0 batch_normalization_91[0][0] \n__________________________________________________________________________________________________\nlambda_31 (Lambda) (None, 1, 1, 1632) 0 swish_91[0][0] \n__________________________________________________________________________________________________\nconv2d_121 (Conv2D) (None, 1, 1, 68) 111044 lambda_31[0][0] \n__________________________________________________________________________________________________\nswish_92 (Swish) (None, 1, 1, 68) 0 conv2d_121[0][0] \n__________________________________________________________________________________________________\nconv2d_122 (Conv2D) (None, 1, 1, 1632) 112608 swish_92[0][0] \n__________________________________________________________________________________________________\nactivation_31 (Activation) (None, 1, 1, 1632) 0 conv2d_122[0][0] \n__________________________________________________________________________________________________\nmultiply_31 (Multiply) (None, 8, 8, 1632) 0 activation_31[0][0] \n swish_91[0][0] \n__________________________________________________________________________________________________\nconv2d_123 (Conv2D) (None, 8, 8, 448) 731136 multiply_31[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_92 (BatchNo (None, 8, 8, 448) 1792 conv2d_123[0][0] \n__________________________________________________________________________________________________\nconv2d_124 (Conv2D) (None, 8, 8, 2688) 1204224 batch_normalization_92[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_93 (BatchNo (None, 8, 8, 2688) 10752 conv2d_124[0][0] \n__________________________________________________________________________________________________\nswish_93 (Swish) (None, 8, 8, 2688) 0 batch_normalization_93[0][0] \n__________________________________________________________________________________________________\ndepthwise_conv2d_32 (DepthwiseC (None, 8, 8, 2688) 24192 swish_93[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_94 (BatchNo (None, 8, 8, 2688) 10752 depthwise_conv2d_32[0][0] \n__________________________________________________________________________________________________\nswish_94 (Swish) (None, 8, 8, 2688) 0 batch_normalization_94[0][0] \n__________________________________________________________________________________________________\nlambda_32 (Lambda) (None, 1, 1, 2688) 0 swish_94[0][0] \n__________________________________________________________________________________________________\nconv2d_125 (Conv2D) (None, 1, 1, 112) 301168 lambda_32[0][0] \n__________________________________________________________________________________________________\nswish_95 (Swish) (None, 1, 1, 112) 0 conv2d_125[0][0] \n__________________________________________________________________________________________________\nconv2d_126 (Conv2D) (None, 1, 1, 2688) 303744 swish_95[0][0] \n__________________________________________________________________________________________________\nactivation_32 (Activation) (None, 1, 1, 2688) 0 conv2d_126[0][0] \n__________________________________________________________________________________________________\nmultiply_32 (Multiply) (None, 8, 8, 2688) 0 activation_32[0][0] \n swish_94[0][0] \n__________________________________________________________________________________________________\nconv2d_127 (Conv2D) (None, 8, 8, 448) 1204224 multiply_32[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_95 (BatchNo (None, 8, 8, 448) 1792 conv2d_127[0][0] \n__________________________________________________________________________________________________\ndrop_connect_25 (DropConnect) (None, 8, 8, 448) 0 batch_normalization_95[0][0] \n__________________________________________________________________________________________________\nadd_25 (Add) (None, 8, 8, 448) 0 drop_connect_25[0][0] \n batch_normalization_92[0][0] \n__________________________________________________________________________________________________\nconv2d_128 (Conv2D) (None, 8, 8, 1792) 802816 add_25[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_96 (BatchNo (None, 8, 8, 1792) 7168 conv2d_128[0][0] \n__________________________________________________________________________________________________\nswish_96 (Swish) (None, 8, 8, 1792) 0 batch_normalization_96[0][0] \n__________________________________________________________________________________________________\nglobal_average_pooling2d_1 (Glo (None, 1792) 0 swish_96[0][0] \n__________________________________________________________________________________________________\nfinal_output (Dense) (None, 1) 1793 global_average_pooling2d_1[0][0] \n==================================================================================================\nTotal params: 17,675,609\nTrainable params: 17,550,409\nNon-trainable params: 125,200\n__________________________________________________________________________________________________\n" ], [ "history = model.fit_generator(generator=train_generator,\n steps_per_epoch=STEP_SIZE_TRAIN,\n validation_data=valid_generator,\n validation_steps=STEP_SIZE_VALID,\n epochs=EPOCHS,\n callbacks=callback_list,\n verbose=2).history", "Epoch 1/20\n - 467s - loss: 0.9229 - acc: 0.3720 - val_loss: 0.4854 - val_acc: 0.5078\nEpoch 2/20\n - 429s - loss: 0.7291 - acc: 0.4226 - val_loss: 0.3720 - val_acc: 0.6990\nEpoch 3/20\n - 429s - loss: 0.6848 - acc: 0.4571 - val_loss: 0.3860 - val_acc: 0.7304\nEpoch 4/20\n - 429s - loss: 0.6829 - acc: 0.4571 - val_loss: 0.4546 - val_acc: 0.6890\nEpoch 5/20\n - 429s - loss: 0.6728 - acc: 0.4615 - val_loss: 0.3843 - val_acc: 0.7061\nEpoch 6/20\n - 429s - loss: 0.6571 - acc: 0.4761 - val_loss: 0.3779 - val_acc: 0.7703\nEpoch 7/20\n - 430s - loss: 0.6442 - acc: 0.4803 - val_loss: 0.4148 - val_acc: 0.6519\nRestoring model weights from the end of the best epoch\nEpoch 00007: early stopping\n" ], [ "fig, ax = plt.subplots(1, 1, sharex='col', figsize=(20, 4))\n\nax.plot(cosine_lr.learning_rates)\nax.set_title('Fine-tune learning rates')\n\nplt.xlabel('Steps')\nplt.ylabel('Learning rate')\nsns.despine()\nplt.show()", "_____no_output_____" ] ], [ [ "# Model loss graph ", "_____no_output_____" ] ], [ [ "plot_metrics(history)", "_____no_output_____" ], [ "# Create empty arays to keep the predictions and labels\ndf_preds = pd.DataFrame(columns=['label', 'pred', 'set'])\ntrain_generator.reset()\nvalid_generator.reset()\n\n# Add train predictions and labels\nfor i in range(STEP_SIZE_TRAIN + 1):\n im, lbl = next(train_generator)\n preds = model.predict(im, batch_size=train_generator.batch_size)\n for index in range(len(preds)):\n df_preds.loc[len(df_preds)] = [lbl[index], preds[index][0], 'train']\n\n# Add validation predictions and labels\nfor i in range(STEP_SIZE_VALID + 1):\n im, lbl = next(valid_generator)\n preds = model.predict(im, batch_size=valid_generator.batch_size)\n for index in range(len(preds)):\n df_preds.loc[len(df_preds)] = [lbl[index], preds[index][0], 'validation']\n\ndf_preds['label'] = df_preds['label'].astype('int')", "_____no_output_____" ], [ "# Classify predictions\ndf_preds['predictions'] = df_preds['pred'].apply(lambda x: classify(x))\n\ntrain_preds = df_preds[df_preds['set'] == 'train']\nvalidation_preds = df_preds[df_preds['set'] == 'validation']", "_____no_output_____" ] ], [ [ "# Model Evaluation", "_____no_output_____" ], [ "## Confusion Matrix\n\n### Original thresholds", "_____no_output_____" ] ], [ [ "plot_confusion_matrix((train_preds['label'], train_preds['predictions']), (validation_preds['label'], validation_preds['predictions']))", "_____no_output_____" ] ], [ [ "## Quadratic Weighted Kappa", "_____no_output_____" ] ], [ [ "evaluate_model((train_preds['label'], train_preds['predictions']), (validation_preds['label'], validation_preds['predictions']))", "Train Cohen Kappa score: 0.738\nValidation Cohen Kappa score: 0.899\nComplete set Cohen Kappa score: 0.746\n" ] ], [ [ "## Apply model to test set and output predictions", "_____no_output_____" ] ], [ [ "preds = apply_tta(model, test_generator, TTA_STEPS)\npredictions = [classify(x) for x in preds]\n\nresults = pd.DataFrame({'id_code':test['id_code'], 'diagnosis':predictions})\nresults['id_code'] = results['id_code'].map(lambda x: str(x)[:-4])", "_____no_output_____" ], [ "# Cleaning created directories\nif os.path.exists(train_dest_path):\n shutil.rmtree(train_dest_path)\nif os.path.exists(validation_dest_path):\n shutil.rmtree(validation_dest_path)\nif os.path.exists(test_dest_path):\n shutil.rmtree(test_dest_path)", "_____no_output_____" ] ], [ [ "# Predictions class distribution", "_____no_output_____" ] ], [ [ "fig = plt.subplots(sharex='col', figsize=(24, 8.7))\nsns.countplot(x=\"diagnosis\", data=results, palette=\"GnBu_d\").set_title('Test')\nsns.despine()\nplt.show()", "_____no_output_____" ], [ "results.to_csv('submission.csv', index=False)\ndisplay(results.head())", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ] ]

[ "CC-BY-3.0" ]

[ "CC-BY-3.0" ]

[ "CC-BY-3.0" ]

[ [ [ "<a href=\"https://colab.research.google.com/github/tastiz/story_scape.html/blob/master/ksStatsPy.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>", "_____no_output_____" ] ], [ [ "# To keep the page organized do all imports here\nfrom sqlalchemy import create_engine\nimport pandas as pd\nfrom scipy import stats\n\n# Database credentials\npostgres_user = 'dabc_student'\npostgres_pw = '7*.8G9QH21'\npostgres_host = '142.93.121.174'\npostgres_port = '5432'\npostgres_db = 'kickstarterprojects'\n\n# use the credentials to start a connection\nengine = create_engine('postgresql://{}:{}@{}:{}/{}'.format(\n postgres_user, postgres_pw, postgres_host, postgres_port, postgres_db))\n\nprojects_df = pd.read_sql_table('ksprojects', con=engine)\n\n# remove the connection\nengine.dispose()\n\n#projects_df.shape\n\n#describes column name and fill tyope\n#projects_df.info()\n\n#projects_df.head(2)\n\n# count the number of unique values in this column\nprojects_df['category'].nunique()\n\n# find the frequency of each value in the column\ncategory_counts = projects_df['category'].value_counts()\n\n# only print the first 10, because 158 are too many to print\n#category_counts.head(10)\n\nd", "_____no_output_____" ] ] ]

[ "markdown", "code" ]

[ [ "markdown" ], [ "code" ] ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ [ [ "# Translate `dzn` to `smt2` for z3", "_____no_output_____" ], [ "### Check Versions of Tools", "_____no_output_____" ] ], [ [ "import os\nimport subprocess\n\nmy_env = os.environ.copy()\noutput = subprocess.check_output(f'''/home/{my_env['USER']}/optimathsat/bin/optimathsat -version''', shell=True, universal_newlines=True)\noutput", "_____no_output_____" ], [ "output = subprocess.check_output(f'''/home/{my_env['USER']}/minizinc/build/minizinc --version''', shell=True, universal_newlines=True)\noutput", "_____no_output_____" ], [ "output = subprocess.check_output(f'''/home/{my_env['USER']}/z3/build/z3 --version''', shell=True, universal_newlines=True)\noutput", "_____no_output_____" ] ], [ [ "First generate the FlatZinc files using the MiniZinc tool. Make sure that a `smt2` folder is located inside `./minizinc/share/minizinc/`. Else, to enable OptiMathSAT's support for global constraints download the [smt2.tar.gz](http://optimathsat.disi.unitn.it/data/smt2.tar.gz) package and unpack it there using\n\n```zsh\ntar xf smt2.tar.gz $MINIZINC_PATH/share/minizinc/\n```\n\nIf next output shows a list of `.mzn` files, then this dependency is satified.", "_____no_output_____" ] ], [ [ "output = subprocess.check_output(f'''ls -la /home/{my_env['USER']}/minizinc/share/minizinc/smt2/''', shell=True, universal_newlines=True)\nprint(output)", "total 292\ndrwxr-xr-x 2 jovyan jovyan 4096 Jan 15 2018 .\ndrwxr-xr-x 11 jovyan jovyan 4096 Jul 11 12:34 ..\n-rw-r--r-- 1 jovyan jovyan 328 Nov 13 2017 alldifferent_except_0.mzn\n-rw-r--r-- 1 jovyan jovyan 382 Nov 13 2017 all_different_int.mzn\n-rw-r--r-- 1 jovyan jovyan 396 Nov 13 2017 all_different_set.mzn\n-rw-r--r-- 1 jovyan jovyan 270 Nov 13 2017 all_disjoint.mzn\n-rw-r--r-- 1 jovyan jovyan 150 Nov 14 2017 all_equal_int.mzn\n-rw-r--r-- 1 jovyan jovyan 164 Nov 13 2017 all_equal_set.mzn\n-rw-r--r-- 1 jovyan jovyan 351 Nov 13 2017 among.mzn\n-rw-r--r-- 1 jovyan jovyan 305 Nov 8 2017 arg_max_float.mzn\n-rw-r--r-- 1 jovyan jovyan 291 Nov 8 2017 arg_max_int.mzn\n-rw-r--r-- 1 jovyan jovyan 306 Nov 8 2017 arg_min_float.mzn\n-rw-r--r-- 1 jovyan jovyan 291 Nov 8 2017 arg_min_int.mzn\n-rw-r--r-- 1 jovyan jovyan 480 Nov 13 2017 at_least_int.mzn\n-rw-r--r-- 1 jovyan jovyan 506 Nov 14 2017 at_least_set.mzn\n-rw-r--r-- 1 jovyan jovyan 340 Nov 13 2017 at_most1.mzn\n-rw-r--r-- 1 jovyan jovyan 474 Nov 13 2017 at_most_int.mzn\n-rw-r--r-- 1 jovyan jovyan 502 Nov 13 2017 at_most_set.mzn\n-rw-r--r-- 1 jovyan jovyan 1162 Nov 8 2017 bin_packing_capa.mzn\n-rw-r--r-- 1 jovyan jovyan 1044 Nov 8 2017 bin_packing_load.mzn\n-rw-r--r-- 1 jovyan jovyan 883 Nov 14 2017 bin_packing.mzn\n-rw-r--r-- 1 jovyan jovyan 765 Nov 14 2017 comparison_rel_array.mzn\n-rw-r--r-- 1 jovyan jovyan 350 Nov 13 2017 count_eq.mzn\n-rw-r--r-- 1 jovyan jovyan 382 Nov 13 2017 count_geq.mzn\n-rw-r--r-- 1 jovyan jovyan 375 Nov 13 2017 count_gt.mzn\n-rw-r--r-- 1 jovyan jovyan 379 Nov 13 2017 count_leq.mzn\n-rw-r--r-- 1 jovyan jovyan 371 Nov 13 2017 count_lt.mzn\n-rw-r--r-- 1 jovyan jovyan 371 Nov 13 2017 count_neq.mzn\n-rw-r--r-- 1 jovyan jovyan 398 Nov 13 2017 decreasing_bool.mzn\n-rw-r--r-- 1 jovyan jovyan 404 Nov 13 2017 decreasing_float.mzn\n-rw-r--r-- 1 jovyan jovyan 393 Nov 13 2017 decreasing_int.mzn\n-rw-r--r-- 1 jovyan jovyan 408 Nov 14 2017 decreasing_set.mzn\n-rw-r--r-- 1 jovyan jovyan 1589 Nov 13 2017 diffn_k.mzn\n-rw-r--r-- 1 jovyan jovyan 853 Nov 14 2017 diffn.mzn\n-rw-r--r-- 1 jovyan jovyan 1731 Nov 13 2017 diffn_nonstrict_k.mzn\n-rw-r--r-- 1 jovyan jovyan 919 Nov 14 2017 diffn_nonstrict.mzn\n-rw-r--r-- 1 jovyan jovyan 276 Nov 14 2017 disjoint.mzn\n-rw-r--r-- 1 jovyan jovyan 836 Nov 8 2017 disjunctive.mzn\n-rw-r--r-- 1 jovyan jovyan 748 Nov 8 2017 disjunctive_strict.mzn\n-rw-r--r-- 1 jovyan jovyan 696 Nov 14 2017 distribute.mzn\n-rw-r--r-- 1 jovyan jovyan 474 Nov 13 2017 exactly_int.mzn\n-rw-r--r-- 1 jovyan jovyan 502 Nov 13 2017 exactly_set.mzn\n-rw-r--r-- 1 jovyan jovyan 851 Nov 14 2017 global_cardinality_closed.mzn\n-rw-r--r-- 1 jovyan jovyan 396 Nov 8 2017 global_cardinality_fn.mzn\n-rw-r--r-- 1 jovyan jovyan 914 Nov 13 2017 global_cardinality_low_up_closed.mzn\n-rw-r--r-- 1 jovyan jovyan 795 Nov 13 2017 global_cardinality_low_up.mzn\n-rw-r--r-- 1 jovyan jovyan 717 Nov 14 2017 global_cardinality.mzn\n-rw-r--r-- 1 jovyan jovyan 398 Nov 13 2017 increasing_bool.mzn\n-rw-r--r-- 1 jovyan jovyan 403 Nov 13 2017 increasing_float.mzn\n-rw-r--r-- 1 jovyan jovyan 394 Nov 13 2017 increasing_int.mzn\n-rw-r--r-- 1 jovyan jovyan 408 Nov 13 2017 increasing_set.mzn\n-rw-r--r-- 1 jovyan jovyan 728 Nov 14 2017 int_set_channel.mzn\n-rw-r--r-- 1 jovyan jovyan 582 Nov 8 2017 inverse.mzn\n-rw-r--r-- 1 jovyan jovyan 827 Nov 8 2017 inverse_set.mzn\n-rw-r--r-- 1 jovyan jovyan 708 Nov 14 2017 link_set_to_booleans.mzn\n-rw-r--r-- 1 jovyan jovyan 375 Nov 11 2017 maximum_float.mzn\n-rw-r--r-- 1 jovyan jovyan 367 Nov 11 2017 maximum_int.mzn\n-rw-r--r-- 1 jovyan jovyan 422 Nov 13 2017 member_bool.mzn\n-rw-r--r-- 1 jovyan jovyan 431 Nov 13 2017 member_float.mzn\n-rw-r--r-- 1 jovyan jovyan 414 Nov 13 2017 member_int.mzn\n-rw-r--r-- 1 jovyan jovyan 442 Nov 13 2017 member_set.mzn\n-rw-r--r-- 1 jovyan jovyan 372 Nov 11 2017 minimum_float.mzn\n-rw-r--r-- 1 jovyan jovyan 367 Nov 11 2017 minimum_int.mzn\n-rw-r--r-- 1 jovyan jovyan 283 Nov 13 2017 nvalue.mzn\n-rw-r--r-- 1 jovyan jovyan 712 Nov 14 2017 range.mzn\n-rw-r--r-- 1 jovyan jovyan 1751 Nov 13 2017 redefinitions-2.0.2.mzn\n-rw-r--r-- 1 jovyan jovyan 1434 Nov 13 2017 redefinitions-2.0.mzn\n-rw-r--r-- 1 jovyan jovyan 678 Nov 15 2017 redefinitions-2.1.mzn\n-rw-r--r-- 1 jovyan jovyan 571 Nov 14 2017 roots.mzn\n-rw-r--r-- 1 jovyan jovyan 764 Nov 8 2017 sum_pred.mzn\n-rw-r--r-- 1 jovyan jovyan 445 Nov 15 2017 symmetric_all_different.mzn\n-rw-r--r-- 1 jovyan jovyan 280 Nov 15 2017 value_precede_int.mzn\n-rw-r--r-- 1 jovyan jovyan 294 Nov 15 2017 value_precede_set.mzn\n\n" ] ], [ [ "## Transform `dzn` to `fzn` Using a `mzn` Model", "_____no_output_____" ], [ "Then transform the desired `.dzn` file to `.fzn` using a `Mz.mzn` MiniZinc model.", "_____no_output_____" ], [ "First list all `dzn` files contained in the `dzn_path` that should get processed.", "_____no_output_____" ] ], [ [ "import os\n\ndzn_files = []\ndzn_path = f'''/home/{my_env['USER']}/data/dzn/'''\n\nfor filename in os.listdir(dzn_path):\n if filename.endswith(\".dzn\"):\n dzn_files.append(filename)\nlen(dzn_files)", "_____no_output_____" ] ], [ [ "#### Model $Mz_1$", "_____no_output_____" ] ], [ [ "import sys\n\nfzn_path = f'''/home/{my_env['USER']}/data/fzn/smt2/Mz1-noAbs/'''\nminizinc_base_cmd = f'''/home/{my_env['USER']}/minizinc/build/minizinc \\\n -Werror \\\n --compile --solver org.minizinc.mzn-fzn \\\n --search-dir /home/{my_env['USER']}/minizinc/share/minizinc/smt2/ \\\n /home/{my_env['USER']}/models/mzn/Mz1-noAbs.mzn '''\ntranslate_count = 0\nfor dzn in dzn_files:\n translate_count += 1\n minizinc_transform_cmd = minizinc_base_cmd + dzn_path + dzn \\\n + ' --output-to-file ' + fzn_path + dzn.replace('.', '-') + '.fzn'\n print(f'''\\r({translate_count}/{len(dzn_files)}) Translating {dzn_path + dzn} to {fzn_path + dzn.replace('.', '-')}.fzn''', end='')\n sys.stdout.flush()\n subprocess.check_output(minizinc_transform_cmd, shell=True, \n universal_newlines=True)", "(278/278) Translating /home/jovyan/data/dzn/R028.dzn to /home/jovyan/data/fzn/smt2/Mz1-noAbs/R028-dzn.fzn" ] ], [ [ "#### Model $Mz_2$", "_____no_output_____" ] ], [ [ "import sys\n\nfzn_path = f'''/home/{my_env['USER']}/data/fzn/smt2/Mz2-noAbs/'''\nminizinc_base_cmd = f'''/home/{my_env['USER']}/minizinc/build/minizinc \\\n -Werror \\\n --compile --solver org.minizinc.mzn-fzn \\\n --search-dir /home/{my_env['USER']}/minizinc/share/minizinc/smt2/ \\\n /home/{my_env['USER']}/models/mzn/Mz2-noAbs.mzn '''\ntranslate_count = 0\nfor dzn in dzn_files:\n translate_count += 1\n minizinc_transform_cmd = minizinc_base_cmd + dzn_path + dzn \\\n + ' --output-to-file ' + fzn_path + dzn.replace('.', '-') + '.fzn'\n print(f'''\\r({translate_count}/{len(dzn_files)}) Translating {dzn_path + dzn} to {fzn_path + dzn.replace('.', '-')}.fzn''', end='')\n sys.stdout.flush()\n subprocess.check_output(minizinc_transform_cmd, shell=True, \n universal_newlines=True)", "(278/278) Translating /home/jovyan/data/dzn/R028.dzn to /home/jovyan/data/fzn/smt2/Mz2-noAbs/R028-dzn.fzn" ] ], [ [ "## Translate `fzn` to `smt2`", "_____no_output_____" ], [ "The generated `.fzn` files can be used to generate a `.smt2` files using the `fzn2smt2.py` script from this [project](https://github.com/PatrickTrentin88/fzn2omt).\n\n**NOTE**: Files `R001` (no cables) and `R002` (one one-sided cable) throw an error while translating.", "_____no_output_____" ], [ "#### $Mz_1$", "_____no_output_____" ] ], [ [ "import os\n\nfzn_files = []\nfzn_path = f'''/home/{my_env['USER']}/data/fzn/smt2/Mz1-noAbs/'''\n\nfor filename in os.listdir(fzn_path):\n if filename.endswith(\".fzn\"):\n fzn_files.append(filename)\nlen(fzn_files)", "_____no_output_____" ], [ "smt2_path = f'''/home/{my_env['USER']}/data/smt2/z3/Mz1-noAbs/'''\nfzn2smt2_base_cmd = f'''/home/{my_env['USER']}/fzn2omt/bin/fzn2z3.py'''\ntranslate_count = 0\nmy_env = os.environ.copy()\nmy_env['PATH'] = f'''/home/{my_env['USER']}/optimathsat/bin/:{my_env['PATH']}'''\nmy_env['PATH'] = f'''/home/{my_env['USER']}/z3/build/:{my_env['PATH']}'''\nfor fzn in fzn_files:\n translate_count += 1\n fzn2smt2_transform_cmd = f'''{fzn2smt2_base_cmd} {fzn_path}{fzn} --smt2 {smt2_path}{fzn.replace('.', '-')}.smt2'''\n print(f'''\\r({translate_count}/{len(fzn_files)}) Translating {fzn_path + fzn} to {smt2_path + fzn.replace('.', '-')}.smt2''', end='')\n try:\n output = subprocess.check_output(fzn2smt2_transform_cmd,\n shell=True,env=my_env, \n universal_newlines=True)\n except Exception as e:\n output = str(e.output)\n print(f'''\\r{output}''', end='')\n sys.stdout.flush()", "(278/278) Translating /home/jovyan/data/fzn/smt2/Mz1-noAbs/R079-dzn.fzn to /home/jovyan/data/smt2/z3/Mz1-noAbs/R079-dzn-fzn.smt2\r" ] ], [ [ "#### $Mz_2$", "_____no_output_____" ] ], [ [ "import os\n\nfzn_files = []\nfzn_path = f'''/home/{my_env['USER']}/data/fzn/smt2/Mz2-noAbs/'''\n\nfor filename in os.listdir(fzn_path):\n if filename.endswith(\".fzn\"):\n fzn_files.append(filename)\nlen(fzn_files)", "_____no_output_____" ], [ "smt2_path = f'''/home/{my_env['USER']}/data/smt2/z3/Mz2-noAbs/'''\nfzn2smt2_base_cmd = f'''/home/{my_env['USER']}/fzn2omt/bin/fzn2z3.py'''\ntranslate_count = 0\nmy_env = os.environ.copy()\nmy_env['PATH'] = f'''/home/{my_env['USER']}/optimathsat/bin/:{my_env['PATH']}'''\nmy_env['PATH'] = f'''/home/{my_env['USER']}/z3/build/:{my_env['PATH']}'''\nfor fzn in fzn_files:\n translate_count += 1\n fzn2smt2_transform_cmd = f'''{fzn2smt2_base_cmd} {fzn_path}{fzn} --smt2 {smt2_path}{fzn.replace('.', '-')}.smt2'''\n print(f'''\\r({translate_count}/{len(fzn_files)}) Translating {fzn_path + fzn} to {smt2_path + fzn.replace('.', '-')}.smt2''', end='')\n try:\n output = subprocess.check_output(fzn2smt2_transform_cmd,\n shell=True,env=my_env, \n universal_newlines=True)\n except Exception as e:\n output = str(e.output)\n print(f'''\\r{output}''', end='')\n sys.stdout.flush()", "(278/278) Translating /home/jovyan/data/fzn/smt2/Mz2-noAbs/R079-dzn.fzn to /home/jovyan/data/smt2/z3/Mz2-noAbs/R079-dzn-fzn.smt2\r" ] ], [ [ "### Adjust `smt2` Files According to Chapter 5.2\n\n- Add lower and upper bounds for the decision variable `pfc`\n- Add number of cavities as comments for later solution extraction (workaround)", "_____no_output_____" ] ], [ [ "import os\nimport re\n\ndef adjust_smt2_file(smt2_path: str, file: str, write_path: str):\n\n with open(smt2_path+'/'+file, 'r+') as myfile:\n data = \"\".join(line for line in myfile)\n\n filename = os.path.splitext(file)[0]\n\n newFile = open(os.path.join(write_path, filename +'.smt2'),\"w+\")\n newFile.write(data)\n newFile.close()\n\n openFile = open(os.path.join(write_path, filename +'.smt2'))\n data = openFile.readlines()\n additionalLines = data[-5:]\n data = data[:-5]\n openFile.close()\n\n newFile = open(os.path.join(write_path, filename +'.smt2'),\"w+\")\n newFile.writelines([item for item in data])\n newFile.close()\n\n with open(os.path.join(write_path, filename +'.smt2'),\"r\") as myfile:\n data = \"\".join(line for line in myfile)\n newFile = open(os.path.join(write_path, filename +'.smt2'),\"w+\")\n matches = re.findall(r'\\(define-fun .\\d\\d \\(\\) Int (\\d+)\\)', data)\n try:\n cavity_count = int(matches[0])\n newFile.write(f''';; k={cavity_count}\\n''')\n newFile.write(f''';; Extract pfc from\\n''')\n for i in range(0,cavity_count):\n newFile.write(f''';; X_INTRODUCED_{str(i)}_\\n''')\n newFile.write(data)\n for i in range(1,cavity_count+1):\n lb = f'''(define-fun lbound{str(i)} () Bool (> X_INTRODUCED_{str(i-1)}_ 0))\\n'''\n ub = f'''(define-fun ubound{str(i)} () Bool (<= X_INTRODUCED_{str(i-1)}_ {str(cavity_count)}))\\n'''\n assertLb = f'''(assert lbound{str(i)})\\n'''\n assertUb = f'''(assert ubound{str(i)})\\n'''\n\n newFile.write(lb)\n newFile.write(ub)\n newFile.write(assertLb)\n newFile.write(assertUb)\n except:\n print(f'''\\nCheck {filename} for completeness - data missing?''')\n newFile.writelines([item for item in additionalLines])\n newFile.close()", "_____no_output_____" ] ], [ [ "#### $Mz_1$", "_____no_output_____" ] ], [ [ "import os\n\nsmt2_files = []\nsmt2_path = f'''/home/{my_env['USER']}/data/smt2/z3/Mz1-noAbs'''\n\nfor filename in os.listdir(smt2_path):\n if filename.endswith(\".smt2\"):\n smt2_files.append(filename)\nlen(smt2_files)", "_____no_output_____" ], [ "fix_count = 0\nfor smt2 in smt2_files:\n fix_count += 1\n print(f'''\\r{fix_count}/{len(smt2_files)} Fixing file {smt2}''', end='')\n adjust_smt2_file(smt2_path=smt2_path, file=smt2, write_path=f'''{smt2_path}''')\n sys.stdout.flush()", "49/278 Fixing file R002-dzn-fzn.smt2\nCheck R002-dzn-fzn for completeness - data missing?\n150/278 Fixing file R001-dzn-fzn.smt2\nCheck R001-dzn-fzn for completeness - data missing?\n278/278 Fixing file R166-dzn-fzn.smt2" ] ], [ [ "#### $Mz_2$", "_____no_output_____" ] ], [ [ "import os\n\nsmt2_files = []\nsmt2_path = f'''/home/{my_env['USER']}/data/smt2/z3/Mz2-noAbs'''\n\nfor filename in os.listdir(smt2_path):\n if filename.endswith(\".smt2\"):\n smt2_files.append(filename)\nlen(smt2_files)", "_____no_output_____" ], [ "fix_count = 0\nfor smt2 in smt2_files:\n fix_count += 1\n print(f'''\\r{fix_count}/{len(smt2_files)} Fixing file {smt2}''', end='')\n adjust_smt2_file(smt2_path=smt2_path, file=smt2, write_path=f'''{smt2_path}''')\n sys.stdout.flush()", "49/278 Fixing file R002-dzn-fzn.smt2\nCheck R002-dzn-fzn for completeness - data missing?\n150/278 Fixing file R001-dzn-fzn.smt2\nCheck R001-dzn-fzn for completeness - data missing?\n278/278 Fixing file R166-dzn-fzn.smt2" ] ], [ [ "## Test Generated `smt2` Files Using `z3`\n\nThis shoud generate the `smt2` file without any error. If this was the case then the `z3` prover can be called on a file by running\n\n\n```zsh\nz3 output/A001-dzn-smt2-fzn.smt2 \n```\n\nyielding something similar to\n\n```zsh\nz3 output/A001-dzn-smt2-fzn.smt2 \nsat\n(objectives\n (obj 41881)\n)\n(model \n (define-fun X_INTRODUCED_981_ () Bool\n false)\n (define-fun X_INTRODUCED_348_ () Bool\n false)\n \n .....\n```", "_____no_output_____" ], [ "#### Test with `smt2` from $Mz_1$", "_____no_output_____" ] ], [ [ "command = f'''/home/{my_env['USER']}/z3/build/z3 /home/{my_env['USER']}/data/smt2/z3/Mz1-noAbs/A001-dzn-fzn.smt2'''\nprint(command)\ntry:\n result = subprocess.check_output(command, shell=True, universal_newlines=True)\nexcept Exception as e:\n print(e.output)\nprint(result)", "_____no_output_____" ] ], [ [ "#### Test with `smt2` from $Mz_2$", "_____no_output_____" ] ], [ [ "result = subprocess.check_output(\n f'''/home/{my_env['USER']}/z3/build/z3 \\\n /home/{my_env['USER']}/data/smt2/z3/Mz2-noAbs/v3/A004-dzn-fzn_v3.smt2''',\n shell=True, universal_newlines=True)\nprint(result)", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "### Cincinnati Salaries\n- https://data.cincinnati-oh.gov/Efficient-Service-Delivery/City-of-Cincinnati-Employees-w-Salaries/wmj4-ygbf", "_____no_output_____" ] ], [ [ "! pip install sodapy\n! pip install pandas", "WARNING: pip is being invoked by an old script wrapper. This will fail in a future version of pip.\nPlease see https://github.com/pypa/pip/issues/5599 for advice on fixing the underlying issue.\nTo avoid this problem you can invoke Python with '-m pip' instead of running pip directly.\nRequirement already satisfied: sodapy in /opt/conda/lib/python3.7/site-packages (2.0.0)\nRequirement already satisfied: requests>=2.20.0 in /opt/conda/lib/python3.7/site-packages (from sodapy) (2.22.0)\nRequirement already satisfied: certifi>=2017.4.17 in /opt/conda/lib/python3.7/site-packages (from requests>=2.20.0->sodapy) (2019.11.28)\nRequirement already satisfied: chardet<3.1.0,>=3.0.2 in /opt/conda/lib/python3.7/site-packages (from requests>=2.20.0->sodapy) (3.0.4)\nRequirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /opt/conda/lib/python3.7/site-packages (from requests>=2.20.0->sodapy) (1.25.7)\nRequirement already satisfied: idna<2.9,>=2.5 in /opt/conda/lib/python3.7/site-packages (from requests>=2.20.0->sodapy) (2.8)\nWARNING: pip is being invoked by an old script wrapper. This will fail in a future version of pip.\nPlease see https://github.com/pypa/pip/issues/5599 for advice on fixing the underlying issue.\nTo avoid this problem you can invoke Python with '-m pip' instead of running pip directly.\nRequirement already satisfied: pandas in /opt/conda/lib/python3.7/site-packages (0.25.3)\nRequirement already satisfied: pytz>=2017.2 in /opt/conda/lib/python3.7/site-packages (from pandas) (2019.3)\nRequirement already satisfied: numpy>=1.13.3 in /opt/conda/lib/python3.7/site-packages (from pandas) (1.17.5)\nRequirement already satisfied: python-dateutil>=2.6.1 in /opt/conda/lib/python3.7/site-packages (from pandas) (2.8.1)\nRequirement already satisfied: six>=1.5 in /opt/conda/lib/python3.7/site-packages (from python-dateutil>=2.6.1->pandas) (1.14.0)\n" ], [ "import pandas as pd\nfrom sodapy import Socrata\n\n# Unauthenticated client only works with public data sets. Note 'None'\n# in place of application token, and no username or password:\nclient = Socrata(\"data.cincinnati-oh.gov\", None)\n\n# Example authenticated client (needed for non-public datasets):\n# client = Socrata(data.cincinnati-oh.gov,\n# MyAppToken,\n# userame=\"[email protected]\",\n# password=\"AFakePassword\")\n\n# First 2000 results, returned as JSON from API / converted to Python list of\n# dictionaries by sodapy.\nresults = client.get(\"wmj4-ygbf\", limit=10000)\n\n# Convert to pandas DataFrame\nresults_df = pd.DataFrame.from_records(results)", "WARNING:root:Requests made without an app_token will be subject to strict throttling limits.\n" ], [ "results_df", "_____no_output_____" ], [ "max(pd.to_numeric(results_df['annual_rt']))", "_____no_output_____" ] ] ]

[ "markdown", "code" ]

[ [ "markdown" ], [ "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "<p><img alt=\"Colaboratory logo\" height=\"45px\" src=\"/img/colab_favicon.ico\" align=\"left\" hspace=\"10px\" vspace=\"0px\"></p>\n\n<h1>Welcome to Colaboratory!</h1>\n\n\nColaboratory is a free Jupyter notebook environment that requires no setup and runs entirely in the cloud.\n\nWith Colaboratory you can write and execute code, save and share your analyses, and access powerful computing resources, all for free from your browser.", "_____no_output_____" ] ], [ [ "#@title Introducing Colaboratory { display-mode: \"form\" }\n#@markdown This 3-minute video gives an overview of the key features of Colaboratory:\nfrom IPython.display import YouTubeVideo\nYouTubeVideo('inN8seMm7UI', width=600, height=400)", "_____no_output_____" ] ], [ [ "## Getting Started\n\nThe document you are reading is a [Jupyter notebook](https://jupyter.org/), hosted in Colaboratory. It is not a static page, but an interactive environment that lets you write and execute code in Python and other languages.\n\nFor example, here is a **code cell** with a short Python script that computes a value, stores it in a variable, and prints the result:", "_____no_output_____" ] ], [ [ "seconds_in_a_day = 24 * 60 * 60\nseconds_in_a_day", "_____no_output_____" ] ], [ [ "To execute the code in the above cell, select it with a click and then either press the play button to the left of the code, or use the keyboard shortcut \"Command/Ctrl+Enter\".\n\nAll cells modify the same global state, so variables that you define by executing a cell can be used in other cells:", "_____no_output_____" ] ], [ [ "seconds_in_a_week = 7 * seconds_in_a_day\nseconds_in_a_week", "_____no_output_____" ] ], [ [ "For more information about working with Colaboratory notebooks, see [Overview of Colaboratory](/notebooks/basic_features_overview.ipynb).\n", "_____no_output_____" ], [ "---\n# Cells\nA notebook is a list of cells. Cells contain either explanatory text or executable code and its output. Click a cell to select it.", "_____no_output_____" ], [ "## Code cells\nBelow is a **code cell**. Once the toolbar button indicates CONNECTED, click in the cell to select it and execute the contents in the following ways:\n\n* Click the **Play icon** in the left gutter of the cell;\n* Type **Cmd/Ctrl+Enter** to run the cell in place;\n* Type **Shift+Enter** to run the cell and move focus to the next cell (adding one if none exists); or\n* Type **Alt+Enter** to run the cell and insert a new code cell immediately below it.\n\nThere are additional options for running some or all cells in the **Runtime** menu.\n", "_____no_output_____" ] ], [ [ "a = 13\na", "_____no_output_____" ] ], [ [ "## Text cells\nThis is a **text cell**. You can **double-click** to edit this cell. Text cells\nuse markdown syntax. To learn more, see our [markdown\nguide](/notebooks/markdown_guide.ipynb).\n\nYou can also add math to text cells using [LaTeX](http://www.latex-project.org/)\nto be rendered by [MathJax](https://www.mathjax.org). Just place the statement\nwithin a pair of **\\$** signs. For example `$\\sqrt{3x-1}+(1+x)^2$` becomes\n$\\sqrt{3x-1}+(1+x)^2.$\n", "_____no_output_____" ], [ "## Adding and moving cells\nYou can add new cells by using the **+ CODE** and **+ TEXT** buttons that show when you hover between cells. These buttons are also in the toolbar above the notebook where they can be used to add a cell below the currently selected cell.\n\nYou can move a cell by selecting it and clicking **Cell Up** or **Cell Down** in the top toolbar. \n\nConsecutive cells can be selected by \"lasso selection\" by dragging from outside one cell and through the group. Non-adjacent cells can be selected concurrently by clicking one and then holding down Ctrl while clicking another. Similarly, using Shift instead of Ctrl will select all intermediate cells.", "_____no_output_____" ], [ "# Integration with Drive\n\nColaboratory is integrated with Google Drive. It allows you to share, comment, and collaborate on the same document with multiple people:\n\n* The **SHARE** button (top-right of the toolbar) allows you to share the notebook and control permissions set on it.\n\n* **File->Make a Copy** creates a copy of the notebook in Drive.\n\n* **File->Save** saves the File to Drive. **File->Save and checkpoint** pins the version so it doesn't get deleted from the revision history. \n\n* **File->Revision history** shows the notebook's revision history. ", "_____no_output_____" ], [ "## Commenting on a cell\nYou can comment on a Colaboratory notebook like you would on a Google Document. Comments are attached to cells, and are displayed next to the cell they refer to. If you have **comment-only** permissions, you will see a comment button on the top right of the cell when you hover over it.\n\nIf you have edit or comment permissions you can comment on a cell in one of three ways: \n\n1. Select a cell and click the comment button in the toolbar above the top-right corner of the cell.\n1. Right click a text cell and select **Add a comment** from the context menu.\n3. Use the shortcut **Ctrl+Shift+M** to add a comment to the currently selected cell. \n\nYou can resolve and reply to comments, and you can target comments to specific collaborators by typing *+[email address]* (e.g., `[email protected]`). Addressed collaborators will be emailed. \n\nThe Comment button in the top-right corner of the page shows all comments attached to the notebook.", "_____no_output_____" ], [ "## More Resources\n- [Guide to Markdown](/notebooks/markdown_guide.ipynb)\n- Colaboratory is built on top of [Jupyter Notebook](https://jupyter.org/).", "_____no_output_____" ], [ "---\n**Original Sources:**\n1. https://colab.research.google.com/notebooks/welcome.ipynb\n2. https://colab.research.google.com/notebooks/basic_features_overview.ipynb", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ] ]

[ "BSD-2-Clause-FreeBSD" ]

[ "BSD-2-Clause-FreeBSD" ]

[ "BSD-2-Clause-FreeBSD" ]

[ [ [ "empty" ] ] ]

[ "empty" ]

[ [ "empty" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "<h1>Table of Contents<span class=\"tocSkip\"></span></h1>\n<div class=\"toc\"><ul class=\"toc-item\"><li><span><a href=\"#Leveraging-Pre-trained-Word-Embedding-for-Text-Classification\" data-toc-modified-id=\"Leveraging-Pre-trained-Word-Embedding-for-Text-Classification-1\"><span class=\"toc-item-num\">1&nbsp;&nbsp;</span>Leveraging Pre-trained Word Embedding for Text Classification</a></span><ul class=\"toc-item\"><li><span><a href=\"#Data-Preparation\" data-toc-modified-id=\"Data-Preparation-1.1\"><span class=\"toc-item-num\">1.1&nbsp;&nbsp;</span>Data Preparation</a></span></li><li><span><a href=\"#Glove\" data-toc-modified-id=\"Glove-1.2\"><span class=\"toc-item-num\">1.2&nbsp;&nbsp;</span>Glove</a></span></li><li><span><a href=\"#Model\" data-toc-modified-id=\"Model-1.3\"><span class=\"toc-item-num\">1.3&nbsp;&nbsp;</span>Model</a></span><ul class=\"toc-item\"><li><span><a href=\"#Model-with-Pretrained-Embedding\" data-toc-modified-id=\"Model-with-Pretrained-Embedding-1.3.1\"><span class=\"toc-item-num\">1.3.1&nbsp;&nbsp;</span>Model with Pretrained Embedding</a></span></li><li><span><a href=\"#Model-without-Pretrained-Embedding\" data-toc-modified-id=\"Model-without-Pretrained-Embedding-1.3.2\"><span class=\"toc-item-num\">1.3.2&nbsp;&nbsp;</span>Model without Pretrained Embedding</a></span></li></ul></li><li><span><a href=\"#Submission\" data-toc-modified-id=\"Submission-1.4\"><span class=\"toc-item-num\">1.4&nbsp;&nbsp;</span>Submission</a></span></li><li><span><a href=\"#Summary\" data-toc-modified-id=\"Summary-1.5\"><span class=\"toc-item-num\">1.5&nbsp;&nbsp;</span>Summary</a></span></li></ul></li><li><span><a href=\"#Reference\" data-toc-modified-id=\"Reference-2\"><span class=\"toc-item-num\">2&nbsp;&nbsp;</span>Reference</a></span></li></ul></div>", "_____no_output_____" ] ], [ [ "# code for loading the format for the notebook\nimport os\n\n# path : store the current path to convert back to it later\npath = os.getcwd()\nos.chdir(os.path.join('..', '..', 'notebook_format'))\n\nfrom formats import load_style\nload_style(plot_style=False)", "_____no_output_____" ], [ "os.chdir(path)\n\n# 1. magic for inline plot\n# 2. magic to print version\n# 3. magic so that the notebook will reload external python modules\n# 4. magic to enable retina (high resolution) plots\n# https://gist.github.com/minrk/3301035\n%matplotlib inline\n%load_ext watermark\n%load_ext autoreload\n%autoreload 2\n%config InlineBackend.figure_format='retina'\n\nimport os\nimport time\nimport numpy as np\nimport pandas as pd\nfrom typing import List, Tuple, Dict\nfrom sklearn.model_selection import train_test_split\nfrom keras import layers\nfrom keras.models import Model\nfrom keras.preprocessing.text import Tokenizer\nfrom keras.utils.np_utils import to_categorical\nfrom keras.preprocessing.sequence import pad_sequences\n\n# prevent scientific notations\npd.set_option('display.float_format', lambda x: '%.3f' % x)\n\n%watermark -a 'Ethen' -d -t -v -p numpy,pandas,sklearn,keras", "Using TensorFlow backend.\n" ] ], [ [ "# Leveraging Pre-trained Word Embedding for Text Classification", "_____no_output_____" ], [ "There are two main ways to obtain word embeddings:\n\n- Learn it from scratch: We specify a neural network architecture and learn the word embeddings jointly with the main task at our hand (e.g. sentiment classification). i.e. we would start off with some random word embeddings, and it would update itself along with the word embeddings.\n- Transfer Learning: The whole idea behind transfer learning is to avoid reinventing the wheel as much as possible. It gives us the capability to transfer knowledge that was gained/learned in some other task and use it to improve the learning of another related task. In practice, one way to do this is for the embedding part of the neural network architecture, we load some other embeddings that were trained on a different machine learning task than the one we are trying to solve and use that to bootstrap the process.\n\nOne area that transfer learning shines is when we have little training data available and using our data alone might not be enough to learn an appropriate task specific embedding/features for our vocabulary. In this case, leveraging a word embedding that captures generic aspect of the language can prove to be beneficial from both a performance and time perspective (i.e. we won't have to spend hours/days training a model from scratch to achieve a similar performance). Keep in mind that, as with all machine learning application, everything is still all about trial and error. What makes a embedding good depends heavily on the task at hand: The word embedding for a movie review sentiment classification model may look very different from a legal document classification model as the semantic of the corpus varies between these two tasks.", "_____no_output_____" ], [ "## Data Preparation", "_____no_output_____" ], [ "We'll use the movie review sentiment analysis dataset from [Kaggle](https://www.kaggle.com/c/word2vec-nlp-tutorial/overview) for this example. It's a binary classification problem with AUC as the ultimate evaluation metric. The next few code chunk performs the usual text preprocessing, build up the word vocabulary and performing a train/test split.", "_____no_output_____" ] ], [ [ "data_dir = 'data'\nsubmission_dir = 'submission'", "_____no_output_____" ], [ "input_path = os.path.join(data_dir, 'word2vec-nlp-tutorial', 'labeledTrainData.tsv')\ndf = pd.read_csv(input_path, delimiter='\\t')\nprint(df.shape)\ndf.head()", "(25000, 3)\n" ], [ "raw_text = df['review'].iloc[0]\nraw_text", "_____no_output_____" ], [ "import re\n\ndef clean_str(string: str) -> str:\n string = re.sub(r\"\\\\\", \"\", string) \n string = re.sub(r\"\\'\", \"\", string) \n string = re.sub(r\"\\\"\", \"\", string) \n return string.strip().lower()", "_____no_output_____" ], [ "from bs4 import BeautifulSoup\n\ndef clean_text(df: pd.DataFrame,\n text_col: str,\n label_col: str) -> Tuple[List[str], List[int]]:\n texts = []\n labels = []\n for raw_text, label in zip(df[text_col], df[label_col]): \n text = BeautifulSoup(raw_text).get_text()\n cleaned_text = clean_str(text)\n texts.append(cleaned_text)\n labels.append(label)\n\n return texts, labels", "_____no_output_____" ], [ "text_col = 'review'\nlabel_col = 'sentiment'\ntexts, labels = clean_text(df, text_col, label_col)\nprint('sample text: ', texts[0])\nprint('corresponding label:', labels[0])", "sample text: with all this stuff going down at the moment with mj ive started listening to his music, watching the odd documentary here and there, watched the wiz and watched moonwalker again. maybe i just want to get a certain insight into this guy who i thought was really cool in the eighties just to maybe make up my mind whether he is guilty or innocent. moonwalker is part biography, part feature film which i remember going to see at the cinema when it was originally released. some of it has subtle messages about mjs feeling towards the press and also the obvious message of drugs are bad mkay.visually impressive but of course this is all about michael jackson so unless you remotely like mj in anyway then you are going to hate this and find it boring. some may call mj an egotist for consenting to the making of this movie but mj and most of his fans would say that he made it for the fans which if true is really nice of him.the actual feature film bit when it finally starts is only on for 20 minutes or so excluding the smooth criminal sequence and joe pesci is convincing as a psychopathic all powerful drug lord. why he wants mj dead so bad is beyond me. because mj overheard his plans? nah, joe pescis character ranted that he wanted people to know it is he who is supplying drugs etc so i dunno, maybe he just hates mjs music.lots of cool things in this like mj turning into a car and a robot and the whole speed demon sequence. also, the director must have had the patience of a saint when it came to filming the kiddy bad sequence as usually directors hate working with one kid let alone a whole bunch of them performing a complex dance scene.bottom line, this movie is for people who like mj on one level or another (which i think is most people). if not, then stay away. it does try and give off a wholesome message and ironically mjs bestest buddy in this movie is a girl! michael jackson is truly one of the most talented people ever to grace this planet but is he guilty? well, with all the attention ive gave this subject....hmmm well i dont know because people can be different behind closed doors, i know this for a fact. he is either an extremely nice but stupid guy or one of the most sickest liars. i hope he is not the latter.\ncorresponding label: 1\n" ], [ "random_state = 1234\nval_split = 0.2\n\nlabels = to_categorical(labels)\ntexts_train, texts_val, y_train, y_val = train_test_split(\n texts, labels,\n test_size=val_split,\n random_state=random_state)\n\nprint('labels shape:', labels.shape)\nprint('train size: ', len(texts_train))\nprint('validation size: ', len(texts_val))", "labels shape: (25000, 2)\ntrain size: 20000\nvalidation size: 5000\n" ], [ "max_num_words = 20000\n\ntokenizer = Tokenizer(num_words=max_num_words, oov_token='<unk>')\ntokenizer.fit_on_texts(texts_train)\nprint('Found %s unique tokens.' % len(tokenizer.word_index))", "Found 74207 unique tokens.\n" ], [ "max_sequence_len = 1000\n\nsequences_train = tokenizer.texts_to_sequences(texts_train)\nx_train = pad_sequences(sequences_train, maxlen=max_sequence_len)\n\nsequences_val = tokenizer.texts_to_sequences(texts_val)\nx_val = pad_sequences(sequences_val, maxlen=max_sequence_len)\n\nsequences_train[0][:5]", "_____no_output_____" ] ], [ [ "## Glove", "_____no_output_____" ], [ "There are many different pretrained word embeddings online. The one we'll be using is from [Glove](https://nlp.stanford.edu/projects/glove/). Others include but not limited to [FastText](https://fasttext.cc/docs/en/crawl-vectors.html), [bpemb](https://github.com/bheinzerling/bpemb).\n\nIf we look at the project's wiki page, we can find any different pretrained embeddings available for us to experiment.\n\n<img src=\"img/pretrained_weights.png\" width=\"100%\" height=\"100%\">", "_____no_output_____" ] ], [ [ "import requests\nfrom tqdm import tqdm\n\ndef download_glove(embedding_type: str='glove.6B.zip'):\n \"\"\"\n download GloVe word vector representations, this step may take a while\n \n Parameters\n ----------\n embedding_type : str, default 'glove.6B.zip'\n Specifying different glove embeddings to download if not already there.\n {'glove.6B.zip', 'glove.42B.300d.zip', 'glove.840B.300d.zip', 'glove.twitter.27B.zip'}\n Be wary of the size. e.g. 'glove.6B.zip' is a 822 MB zipped, 2GB unzipped\n \"\"\"\n\n base_url = 'http://nlp.stanford.edu/data/'\n if not os.path.isfile(embedding_type):\n url = base_url + embedding_type\n\n # the following section is a pretty generic http get request for\n # saving large files, provides progress bars for checking progress\n response = requests.get(url, stream=True)\n response.raise_for_status()\n\n content_len = response.headers.get('Content-Length')\n total = int(content_len) if content_len is not None else 0\n\n with tqdm(unit='B', total=total) as pbar, open(embedding_type, 'wb') as f:\n for chunk in response.iter_content(chunk_size=1024):\n if chunk:\n pbar.update(len(chunk))\n f.write(chunk)\n\n if response.headers.get('Content-Type') == 'application/zip':\n from zipfile import ZipFile\n with ZipFile(embedding_type, 'r') as f:\n f.extractall(embedding_type.strip('.zip'))\n\n\ndownload_glove()", "_____no_output_____" ] ], [ [ "The way we'll leverage the pretrained embedding is to first read it in as a dictionary lookup, where the key is the word and the value is the corresponding word embedding. Then for each token in our vocabulary, we'll lookup this dictionary to see if there's a pretrained embedding available, if there is, we'll use the pretrained embedding, if there isn't, we'll leave the embedding for this word in its original randomly initialized form.\n\nThe format for this particular pretrained embedding is for every line, we have a space delimited values, where the first token is the word, and the rest are its corresponding embedding values. e.g. the first line from the line looks like:\n\n```\nthe -0.038194 -0.24487 0.72812 -0.39961 0.083172 0.043953 -0.39141 0.3344 -0.57545 0.087459 0.28787 -0.06731 0.30906 -0.26384 -0.13231 -0.20757 0.33395 -0.33848 -0.31743 -0.48336 0.1464 -0.37304 0.34577 0.052041 0.44946 -0.46971 0.02628 -0.54155 -0.15518 -0.14107 -0.039722 0.28277 0.14393 0.23464 -0.31021 0.086173 0.20397 0.52624 0.17164 -0.082378 -0.71787 -0.41531 0.20335 -0.12763 0.41367 0.55187 0.57908 -0.33477 -0.36559 -0.54857 -0.062892 0.26584 0.30205 0.99775 -0.80481 -3.0243 0.01254 -0.36942 2.2167 0.72201 -0.24978 0.92136 0.034514 0.46745 1.1079 -0.19358 -0.074575 0.23353 -0.052062 -0.22044 0.057162 -0.15806 -0.30798 -0.41625 0.37972 0.15006 -0.53212 -0.2055 -1.2526 0.071624 0.70565 0.49744 -0.42063 0.26148 -1.538 -0.30223 -0.073438 -0.28312 0.37104 -0.25217 0.016215 -0.017099 -0.38984 0.87424 -0.72569 -0.51058 -0.52028 -0.1459 0.8278 0.27062\n```", "_____no_output_____" ] ], [ [ "def get_embedding_lookup(embedding_path) -> Dict[str, np.ndarray]:\n embedding_lookup = {}\n with open(embedding_path) as f:\n for line in f:\n values = line.split()\n word = values[0]\n coef = np.array(values[1:], dtype=np.float32)\n embedding_lookup[word] = coef\n\n return embedding_lookup\n\n\ndef get_pretrained_embedding(embedding_path: str,\n index2word: Dict[int, str],\n max_features: int) -> np.ndarray:\n embedding_lookup = get_embedding_lookup(embedding_path)\n\n pretrained_embedding = np.stack(list(embedding_lookup.values()))\n embedding_dim = pretrained_embedding.shape[1]\n embeddings = np.random.normal(pretrained_embedding.mean(),\n pretrained_embedding.std(),\n (max_features, embedding_dim)).astype(np.float32)\n # we track how many tokens in our vocabulary exists in the pre-trained embedding,\n # i.e. how many tokens has a pre-trained embedding from this particular file\n n_found = 0\n \n # the loop starts from 1 due to keras' Tokenizer reserves 0 for padding index\n for i in range(1, max_features):\n word = index2word[i]\n embedding_vector = embedding_lookup.get(word)\n if embedding_vector is not None:\n embeddings[i] = embedding_vector\n n_found += 1\n\n print('number of words found:', n_found)\n return embeddings", "_____no_output_____" ], [ "glove_path = os.path.join('glove.6B', 'glove.6B.100d.txt')\nmax_features = max_num_words + 1\n\npretrained_embedding = get_pretrained_embedding(glove_path, tokenizer.index_word, max_features)\npretrained_embedding.shape", "number of words found: 19654\n" ] ], [ [ "## Model", "_____no_output_____" ], [ "To train our text classifier, we specify a 1D convolutional network. Our embedding layer can either be initialized randomly or loaded from a pre-trained embedding. Note that for the pre-trained embedding case, apart from loading the weights, we also \"freeze\" the embedding layer, i.e. we set its trainable attribute to False. This idea is often times used in transfer learning, where when parts of a model are pre-trained (in our case, only our Embedding layer), and parts of it are randomly initialized, the pre-trained part should ideally not be trained together with the randomly initialized part. The rationale behind it is that a large gradient update triggered by the randomly initialized layer would become very disruptive to those pre-trained weights.\n\nOnce we train the randomly initialized weights for a few iterations, we can then go about un-freezing the layers that were loaded with pre-trained weights, and do an update on the weight for the entire thing. The [keras documentation](https://keras.io/applications/#fine-tune-inceptionv3-on-a-new-set-of-classes) also provides an example of how to do this, although the example is for image models, the same idea can also be applied here, and can be something that's worth experimenting.", "_____no_output_____" ] ], [ [ "def simple_text_cnn(max_sequence_len: int,\n max_features: int,\n num_classes: int,\n optimizer: str='adam',\n metrics: List[str]=['acc'],\n pretrained_embedding: np.ndarray=None) -> Model:\n\n sequence_input = layers.Input(shape=(max_sequence_len,), dtype='int32')\n if pretrained_embedding is None:\n embedded_sequences = layers.Embedding(max_features, 100,\n name='embedding')(sequence_input)\n else:\n embedded_sequences = layers.Embedding(max_features, pretrained_embedding.shape[1],\n weights=[pretrained_embedding],\n name='embedding',\n trainable=False)(sequence_input)\n\n conv1 = layers.Conv1D(128, 5, activation='relu')(embedded_sequences)\n pool1 = layers.MaxPooling1D(5)(conv1)\n conv2 = layers.Conv1D(128, 5, activation='relu')(pool1)\n pool2 = layers.MaxPooling1D(5)(conv2)\n conv3 = layers.Conv1D(128, 5, activation='relu')(pool2)\n pool3 = layers.MaxPooling1D(35)(conv3)\n flatten = layers.Flatten()(pool3)\n dense = layers.Dense(128, activation='relu')(flatten)\n preds = layers.Dense(num_classes, activation='softmax')(dense)\n\n model = Model(sequence_input, preds)\n model.compile(loss='categorical_crossentropy',\n optimizer=optimizer,\n metrics=metrics)\n return model", "_____no_output_____" ] ], [ [ "### Model with Pretrained Embedding", "_____no_output_____" ] ], [ [ "num_classes = 2\nmodel1 = simple_text_cnn(max_sequence_len, max_features, num_classes,\n pretrained_embedding=pretrained_embedding)\nmodel1.summary()", "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:66: The name tf.get_default_graph is deprecated. Please use tf.compat.v1.get_default_graph instead.\n\nWARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:541: The name tf.placeholder is deprecated. Please use tf.compat.v1.placeholder instead.\n\nWARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:4432: The name tf.random_uniform is deprecated. Please use tf.random.uniform instead.\n\nWARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:190: The name tf.get_default_session is deprecated. Please use tf.compat.v1.get_default_session instead.\n\nWARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:197: The name tf.ConfigProto is deprecated. Please use tf.compat.v1.ConfigProto instead.\n\nWARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:203: The name tf.Session is deprecated. Please use tf.compat.v1.Session instead.\n\nWARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:207: The name tf.global_variables is deprecated. Please use tf.compat.v1.global_variables instead.\n\nWARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:216: The name tf.is_variable_initialized is deprecated. Please use tf.compat.v1.is_variable_initialized instead.\n\nWARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:223: The name tf.variables_initializer is deprecated. Please use tf.compat.v1.variables_initializer instead.\n\nWARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:4267: The name tf.nn.max_pool is deprecated. Please use tf.nn.max_pool2d instead.\n\nWARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/optimizers.py:793: The name tf.train.Optimizer is deprecated. Please use tf.compat.v1.train.Optimizer instead.\n\nWARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:3576: The name tf.log is deprecated. Please use tf.math.log instead.\n\nModel: \"model_1\"\n_________________________________________________________________\nLayer (type) Output Shape Param # \n=================================================================\ninput_1 (InputLayer) (None, 1000) 0 \n_________________________________________________________________\nembedding (Embedding) (None, 1000, 100) 2000100 \n_________________________________________________________________\nconv1d_1 (Conv1D) (None, 996, 128) 64128 \n_________________________________________________________________\nmax_pooling1d_1 (MaxPooling1 (None, 199, 128) 0 \n_________________________________________________________________\nconv1d_2 (Conv1D) (None, 195, 128) 82048 \n_________________________________________________________________\nmax_pooling1d_2 (MaxPooling1 (None, 39, 128) 0 \n_________________________________________________________________\nconv1d_3 (Conv1D) (None, 35, 128) 82048 \n_________________________________________________________________\nmax_pooling1d_3 (MaxPooling1 (None, 1, 128) 0 \n_________________________________________________________________\nflatten_1 (Flatten) (None, 128) 0 \n_________________________________________________________________\ndense_1 (Dense) (None, 128) 16512 \n_________________________________________________________________\ndense_2 (Dense) (None, 2) 258 \n=================================================================\nTotal params: 2,245,094\nTrainable params: 244,994\nNon-trainable params: 2,000,100\n_________________________________________________________________\n" ] ], [ [ "We can confirm whether our embedding layer is trainable by looping through each layer and checking the trainable attribute.", "_____no_output_____" ] ], [ [ "df_model_layers = pd.DataFrame(\n [(layer.name, layer.trainable, layer.count_params()) for layer in model1.layers],\n columns=['layer', 'trainable', 'n_params']\n)\ndf_model_layers", "_____no_output_____" ], [ "# time : 70\n# test performance : auc 0.93212\nstart = time.time()\nhistory1 = model1.fit(x_train, y_train,\n validation_data=(x_val, y_val),\n batch_size=128,\n epochs=8)\nend = time.time()\nelapse1 = end - start\nelapse1", "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow_core/python/ops/math_grad.py:1424: where (from tensorflow.python.ops.array_ops) is deprecated and will be removed in a future version.\nInstructions for updating:\nUse tf.where in 2.0, which has the same broadcast rule as np.where\nWARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:1033: The name tf.assign_add is deprecated. Please use tf.compat.v1.assign_add instead.\n\nWARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:1020: The name tf.assign is deprecated. Please use tf.compat.v1.assign instead.\n\nTrain on 20000 samples, validate on 5000 samples\nEpoch 1/8\n20000/20000 [==============================] - 12s 604us/step - loss: 0.5854 - acc: 0.6748 - val_loss: 0.4772 - val_acc: 0.7808\nEpoch 2/8\n20000/20000 [==============================] - 8s 416us/step - loss: 0.4001 - acc: 0.8186 - val_loss: 0.3766 - val_acc: 0.8352\nEpoch 3/8\n20000/20000 [==============================] - 8s 414us/step - loss: 0.3428 - acc: 0.8507 - val_loss: 0.4276 - val_acc: 0.7966\nEpoch 4/8\n20000/20000 [==============================] - 8s 415us/step - loss: 0.2790 - acc: 0.8842 - val_loss: 0.3433 - val_acc: 0.8594\nEpoch 5/8\n20000/20000 [==============================] - 8s 415us/step - loss: 0.2469 - acc: 0.8987 - val_loss: 0.4015 - val_acc: 0.8310\nEpoch 6/8\n20000/20000 [==============================] - 8s 420us/step - loss: 0.1782 - acc: 0.9289 - val_loss: 0.4670 - val_acc: 0.8296\nEpoch 7/8\n20000/20000 [==============================] - 8s 419us/step - loss: 0.1017 - acc: 0.9643 - val_loss: 0.5965 - val_acc: 0.8146\nEpoch 8/8\n20000/20000 [==============================] - 8s 418us/step - loss: 0.0680 - acc: 0.9758 - val_loss: 0.6876 - val_acc: 0.8332\n" ] ], [ [ "### Model without Pretrained Embedding", "_____no_output_____" ] ], [ [ "num_classes = 2\nmodel2 = simple_text_cnn(max_sequence_len, max_features, num_classes)\nmodel2.summary()", "Model: \"model_2\"\n_________________________________________________________________\nLayer (type) Output Shape Param # \n=================================================================\ninput_2 (InputLayer) (None, 1000) 0 \n_________________________________________________________________\nembedding (Embedding) (None, 1000, 100) 2000100 \n_________________________________________________________________\nconv1d_4 (Conv1D) (None, 996, 128) 64128 \n_________________________________________________________________\nmax_pooling1d_4 (MaxPooling1 (None, 199, 128) 0 \n_________________________________________________________________\nconv1d_5 (Conv1D) (None, 195, 128) 82048 \n_________________________________________________________________\nmax_pooling1d_5 (MaxPooling1 (None, 39, 128) 0 \n_________________________________________________________________\nconv1d_6 (Conv1D) (None, 35, 128) 82048 \n_________________________________________________________________\nmax_pooling1d_6 (MaxPooling1 (None, 1, 128) 0 \n_________________________________________________________________\nflatten_2 (Flatten) (None, 128) 0 \n_________________________________________________________________\ndense_3 (Dense) (None, 128) 16512 \n_________________________________________________________________\ndense_4 (Dense) (None, 2) 258 \n=================================================================\nTotal params: 2,245,094\nTrainable params: 2,245,094\nNon-trainable params: 0\n_________________________________________________________________\n" ], [ "# time : 86 secs\n# test performance : auc 0.92310\nstart = time.time()\nhistory1 = model2.fit(x_train, y_train,\n validation_data=(x_val, y_val),\n batch_size=128,\n epochs=8)\nend = time.time()\nelapse1 = end - start\nelapse1", "Train on 20000 samples, validate on 5000 samples\nEpoch 1/8\n20000/20000 [==============================] - 11s 570us/step - loss: 0.5010 - acc: 0.7065 - val_loss: 0.3016 - val_acc: 0.8730\nEpoch 2/8\n20000/20000 [==============================] - 11s 542us/step - loss: 0.2024 - acc: 0.9243 - val_loss: 0.2816 - val_acc: 0.8824\nEpoch 3/8\n20000/20000 [==============================] - 11s 538us/step - loss: 0.0806 - acc: 0.9734 - val_loss: 0.3552 - val_acc: 0.8812\nEpoch 4/8\n20000/20000 [==============================] - 11s 535us/step - loss: 0.0272 - acc: 0.9917 - val_loss: 0.4671 - val_acc: 0.8836\nEpoch 5/8\n20000/20000 [==============================] - 11s 543us/step - loss: 0.0088 - acc: 0.9973 - val_loss: 0.6534 - val_acc: 0.8788\nEpoch 6/8\n20000/20000 [==============================] - 11s 542us/step - loss: 0.0090 - acc: 0.9973 - val_loss: 0.7522 - val_acc: 0.8740\nEpoch 7/8\n20000/20000 [==============================] - 11s 542us/step - loss: 0.0104 - acc: 0.9967 - val_loss: 1.0453 - val_acc: 0.8480\nEpoch 8/8\n20000/20000 [==============================] - 11s 543us/step - loss: 0.0205 - acc: 0.9924 - val_loss: 0.6930 - val_acc: 0.8712\n" ] ], [ [ "## Submission", "_____no_output_____" ], [ "For the submission section, we read in and preprocess the test data provided by the competition, then generate the predicted probability column for both the model that uses pretrained embedding and one that doesn't to compare their performance.", "_____no_output_____" ] ], [ [ "input_path = os.path.join(data_dir, 'word2vec-nlp-tutorial', 'testData.tsv')\ndf_test = pd.read_csv(input_path, delimiter='\\t')\nprint(df_test.shape)\ndf_test.head()", "(25000, 2)\n" ], [ "def clean_text_without_label(df: pd.DataFrame, text_col: str) -> List[str]:\n texts = []\n for raw_text in df[text_col]:\n text = BeautifulSoup(raw_text).get_text()\n cleaned_text = clean_str(text)\n texts.append(cleaned_text)\n\n return texts", "_____no_output_____" ], [ "texts_test = clean_text_without_label(df_test, text_col)\nsequences_test = tokenizer.texts_to_sequences(texts_test)\nx_test = pad_sequences(sequences_test, maxlen=max_sequence_len)\nlen(x_test)", "_____no_output_____" ], [ "def create_submission(ids, predictions, ids_col, label_col, submission_path) -> pd.DataFrame:\n df_submission = pd.DataFrame({\n ids_col: ids,\n label_col: predictions\n }, columns=[ids_col, label_col])\n\n if submission_path is not None:\n # create the directory if need be, e.g. if the submission_path = submission/submission.csv\n # we'll create the submission directory first if it doesn't exist\n directory = os.path.split(submission_path)[0]\n if (directory != '' or directory != '.') and not os.path.isdir(directory):\n os.makedirs(directory, exist_ok=True)\n\n df_submission.to_csv(submission_path, index=False, header=True)\n\n return df_submission", "_____no_output_____" ], [ "ids_col = 'id'\nlabel_col = 'sentiment'\nids = df_test[ids_col]\n\nmodels = {\n 'pretrained_embedding': model1,\n 'without_pretrained_embedding': model2\n}\n\nfor model_name, model in models.items():\n print('generating submission for: ', model_name)\n submission_path = os.path.join(submission_dir, '{}_submission.csv'.format(model_name))\n predictions = model.predict(x_test, verbose=1)[:, 1]\n df_submission = create_submission(ids, predictions, ids_col, label_col, submission_path)\n\n# sanity check to make sure the size and the output of the submission makes sense\nprint(df_submission.shape)\ndf_submission.head()", "generating submission for: pretrained_embedding\n25000/25000 [==============================] - 6s 228us/step\ngenerating submission for: without_pretrained_embedding\n25000/25000 [==============================] - 6s 222us/step\n(25000, 2)\n" ] ], [ [ "## Summary", "_____no_output_____" ], [ "In this article, we took a look at how to leverage pre-trained word embeddings for our text classification task. There're also various Kaggle Kernels [here](https://www.kaggle.com/sudalairajkumar/a-look-at-different-embeddings) and [here](https://www.kaggle.com/sbongo/do-pretrained-embeddings-give-you-the-extra-edge) that experiments whether different pre-trained embeddings or even an ensemble of models each with a different pre-trained embedding on various text classification tasks to see if it gives us an edge. ", "_____no_output_____" ], [ "# Reference", "_____no_output_____" ], [ "- [Blog: Text Classification, Part I - Convolutional Networks](https://richliao.github.io/supervised/classification/2016/11/26/textclassifier-convolutional/)\n- [Blog: Using pre-trained word embeddings in a Keras model](https://blog.keras.io/using-pre-trained-word-embeddings-in-a-keras-model.html)\n- [Jupyter Notebook - Deep Learning with Python - Using Word Embeddings](https://nbviewer.jupyter.org/github/fchollet/deep-learning-with-python-notebooks/blob/master/6.1-using-word-embeddings.ipynb)", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown", "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown", "markdown", "markdown", "markdown" ] ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ [ [ "<p><font size=\"6\"><b>04 - Pandas: Working with time series data</b></font></p>\n\n> *© 2021, Joris Van den Bossche and Stijn Van Hoey (<mailto:[email protected]>, <mailto:[email protected]>). Licensed under [CC BY 4.0 Creative Commons](http://creativecommons.org/licenses/by/4.0/)*\n\n---", "_____no_output_____" ] ], [ [ "import pandas as pd\nimport numpy as np\nimport matplotlib.pyplot as plt\n\nplt.style.use('ggplot')", "_____no_output_____" ] ], [ [ "# Introduction: `datetime` module", "_____no_output_____" ], [ "Standard Python contains the `datetime` module to handle date and time data:", "_____no_output_____" ] ], [ [ "import datetime", "_____no_output_____" ], [ "dt = datetime.datetime(year=2016, month=12, day=19, hour=13, minute=30)\ndt", "_____no_output_____" ], [ "print(dt) # .day,...", "_____no_output_____" ], [ "print(dt.strftime(\"%d %B %Y\"))", "_____no_output_____" ] ], [ [ "# Dates and times in pandas", "_____no_output_____" ], [ "## The ``Timestamp`` object", "_____no_output_____" ], [ "Pandas has its own date and time objects, which are compatible with the standard `datetime` objects, but provide some more functionality to work with. \n\nThe `Timestamp` object can also be constructed from a string:", "_____no_output_____" ] ], [ [ "ts = pd.Timestamp('2016-12-19')\nts", "_____no_output_____" ] ], [ [ "Like with `datetime.datetime` objects, there are several useful attributes available on the `Timestamp`. For example, we can get the month (experiment with tab completion!):", "_____no_output_____" ] ], [ [ "ts.month", "_____no_output_____" ] ], [ [ "There is also a `Timedelta` type, which can e.g. be used to add intervals of time:", "_____no_output_____" ] ], [ [ "ts + pd.Timedelta('5 days')", "_____no_output_____" ] ], [ [ "## Parsing datetime strings", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "Unfortunately, when working with real world data, you encounter many different `datetime` formats. Most of the time when you have to deal with them, they come in text format, e.g. from a `CSV` file. To work with those data in Pandas, we first have to *parse* the strings to actual `Timestamp` objects.", "_____no_output_____" ], [ "<div class=\"alert alert-info\">\n<b>REMEMBER</b>: <br><br>\n\nTo convert string formatted dates to Timestamp objects: use the `pandas.to_datetime` function\n\n</div>", "_____no_output_____" ] ], [ [ "pd.to_datetime(\"2016-12-09\")", "_____no_output_____" ], [ "pd.to_datetime(\"09/12/2016\")", "_____no_output_____" ], [ "pd.to_datetime(\"09/12/2016\", dayfirst=True)", "_____no_output_____" ], [ "pd.to_datetime(\"09/12/2016\", format=\"%d/%m/%Y\")", "_____no_output_____" ] ], [ [ "A detailed overview of how to specify the `format` string, see the table in the python documentation: https://docs.python.org/3/library/datetime.html#strftime-and-strptime-behavior", "_____no_output_____" ], [ "## `Timestamp` data in a Series or DataFrame column", "_____no_output_____" ] ], [ [ "s = pd.Series(['2016-12-09 10:00:00', '2016-12-09 11:00:00', '2016-12-09 12:00:00'])", "_____no_output_____" ], [ "s", "_____no_output_____" ] ], [ [ "The `to_datetime` function can also be used to convert a full series of strings:", "_____no_output_____" ] ], [ [ "ts = pd.to_datetime(s)", "_____no_output_____" ], [ "ts", "_____no_output_____" ] ], [ [ "Notice the data type of this series has changed: the `datetime64[ns]` dtype. This indicates that we have a series of actual datetime values.", "_____no_output_____" ], [ "The same attributes as on single `Timestamp`s are also available on a Series with datetime data, using the **`.dt`** accessor:", "_____no_output_____" ] ], [ [ "ts.dt.hour", "_____no_output_____" ], [ "ts.dt.dayofweek", "_____no_output_____" ] ], [ [ "To quickly construct some regular time series data, the [``pd.date_range``](http://pandas.pydata.org/pandas-docs/stable/generated/pandas.date_range.html) function comes in handy:", "_____no_output_____" ] ], [ [ "pd.Series(pd.date_range(start=\"2016-01-01\", periods=10, freq='3H'))", "_____no_output_____" ] ], [ [ "# Time series data: `Timestamp` in the index", "_____no_output_____" ], [ "## River discharge example data", "_____no_output_____" ], [ "For the following demonstration of the time series functionality, we use a sample of discharge data of the Maarkebeek (Flanders) with 3 hour averaged values, derived from the [Waterinfo website](https://www.waterinfo.be/).", "_____no_output_____" ] ], [ [ "data = pd.read_csv(\"data/vmm_flowdata.csv\")", "_____no_output_____" ], [ "data.head()", "_____no_output_____" ] ], [ [ "We already know how to parse a date column with Pandas:", "_____no_output_____" ] ], [ [ "data['Time'] = pd.to_datetime(data['Time'])", "_____no_output_____" ] ], [ [ "With `set_index('datetime')`, we set the column with datetime values as the index, which can be done by both `Series` and `DataFrame`.", "_____no_output_____" ] ], [ [ "data = data.set_index(\"Time\")", "_____no_output_____" ], [ "data", "_____no_output_____" ] ], [ [ "The steps above are provided as built-in functionality of `read_csv`:", "_____no_output_____" ] ], [ [ "data = pd.read_csv(\"data/vmm_flowdata.csv\", index_col=0, parse_dates=True)", "_____no_output_____" ] ], [ [ "<div class=\"alert alert-info\">\n<b>REMEMBER</b>: <br><br>\n\n`pd.read_csv` provides a lot of built-in functionality to support this kind of transactions when reading in a file! Check the help of the read_csv function...\n\n</div>", "_____no_output_____" ], [ "## The DatetimeIndex", "_____no_output_____" ], [ "When we ensure the DataFrame has a `DatetimeIndex`, time-series related functionality becomes available:", "_____no_output_____" ] ], [ [ "data.index", "_____no_output_____" ] ], [ [ "Similar to a Series with datetime data, there are some attributes of the timestamp values available:", "_____no_output_____" ] ], [ [ "data.index.day", "_____no_output_____" ], [ "data.index.dayofyear", "_____no_output_____" ], [ "data.index.year", "_____no_output_____" ] ], [ [ "The `plot` method will also adapt its labels (when you zoom in, you can see the different levels of detail of the datetime labels):", "_____no_output_____" ] ], [ [ "%matplotlib widget", "_____no_output_____" ], [ "data.plot()", "_____no_output_____" ], [ "# switching back to static inline plots (the default)\n%matplotlib inline", "_____no_output_____" ] ], [ [ "We have too much data to sensibly plot on one figure. Let's see how we can easily select part of the data or aggregate the data to other time resolutions in the next sections.", "_____no_output_____" ], [ "## Selecting data from a time series", "_____no_output_____" ], [ "We can use label based indexing on a timeseries as expected:", "_____no_output_____" ] ], [ [ "data[pd.Timestamp(\"2012-01-01 09:00\"):pd.Timestamp(\"2012-01-01 19:00\")]", "_____no_output_____" ] ], [ [ "But, for convenience, indexing a time series also works with strings:", "_____no_output_____" ] ], [ [ "data[\"2012-01-01 09:00\":\"2012-01-01 19:00\"]", "_____no_output_____" ] ], [ [ "A nice feature is **\"partial string\" indexing**, where we can do implicit slicing by providing a partial datetime string.\n\nE.g. all data of 2013:", "_____no_output_____" ] ], [ [ "data['2013':]", "_____no_output_____" ] ], [ [ "Or all data of January up to March 2012:", "_____no_output_____" ] ], [ [ "data['2012-01':'2012-03']", "_____no_output_____" ] ], [ [ "<div class=\"alert alert-success\">\n\n<b>EXERCISE</b>:\n\n <ul>\n <li>select all data starting from 2012</li>\n</ul>\n</div>", "_____no_output_____" ] ], [ [ "data['2012':]", "_____no_output_____" ] ], [ [ "<div class=\"alert alert-success\">\n\n<b>EXERCISE</b>:\n\n <ul>\n <li>select all data in January for all different years</li>\n</ul>\n</div>", "_____no_output_____" ] ], [ [ "data[data.index.month == 1]", "_____no_output_____" ] ], [ [ "<div class=\"alert alert-success\">\n\n<b>EXERCISE</b>:\n\n <ul>\n <li>select all data in April, May and June for all different years</li>\n</ul>\n</div>", "_____no_output_____" ] ], [ [ "data[data.index.month.isin([4, 5, 6])]", "_____no_output_____" ] ], [ [ "<div class=\"alert alert-success\">\n\n<b>EXERCISE</b>:\n\n <ul>\n <li>select all 'daytime' data (between 8h and 20h) for all days</li>\n</ul>\n</div>", "_____no_output_____" ] ], [ [ "data[(data.index.hour > 8) & (data.index.hour < 20)]", "_____no_output_____" ] ], [ [ "## The power of pandas: `resample`", "_____no_output_____" ], [ "A very powerfull method is **`resample`: converting the frequency of the time series** (e.g. from hourly to daily data).\n\nThe time series has a frequency of 1 hour. I want to change this to daily:", "_____no_output_____" ] ], [ [ "data.resample('D').mean().head()", "_____no_output_____" ] ], [ [ "Other mathematical methods can also be specified:", "_____no_output_____" ] ], [ [ "data.resample('D').max().head()", "_____no_output_____" ] ], [ [ "<div class=\"alert alert-info\">\n<b>REMEMBER</b>: <br><br>\n\nThe string to specify the new time frequency: http://pandas.pydata.org/pandas-docs/stable/user_guide/timeseries.html#offset-aliases <br>\n\nThese strings can also be combined with numbers, eg `'10D'`...\n\n</div>", "_____no_output_____" ] ], [ [ "data.resample('M').mean().plot() # 10D", "_____no_output_____" ] ], [ [ "<div class=\"alert alert-success\">\n\n<b>EXERCISE</b>:\n\n <ul>\n <li>Plot the monthly standard deviation of the columns</li>\n</ul>\n</div>", "_____no_output_____" ] ], [ [ "data.resample('M').std().plot() # 'A'", "_____no_output_____" ] ], [ [ "<div class=\"alert alert-success\">\n\n<b>EXERCISE</b>:\n\n <ul>\n <li>Plot the monthly mean and median values for the years 2011-2012 for 'L06_347'<br><br></li>\n</ul>\n\n__Note__ Did you know <a href=\"https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.agg.html\"><code>agg</code></a> to derive multiple statistics at the same time?\n\n</div>", "_____no_output_____" ] ], [ [ "subset = data['2011':'2012']['L06_347']\nsubset.resample('M').agg(['mean', 'median']).plot()", "_____no_output_____" ] ], [ [ "<div class=\"alert alert-success\">\n\n<b>EXERCISE</b>:\n\n <ul>\n <li>plot the monthly mininum and maximum daily average value of the 'LS06_348' column</li>\n</ul>\n</div>", "_____no_output_____" ] ], [ [ "daily = data['LS06_348'].resample('D').mean() # daily averages calculated", "_____no_output_____" ], [ "daily.resample('M').agg(['min', 'max']).plot() # monthly minimum and maximum values of these daily averages", "_____no_output_____" ] ], [ [ "<div class=\"alert alert-success\">\n<b>EXERCISE</b>:\n\n <ul>\n <li>Make a bar plot of the mean of the stations in year of 2013</li>\n</ul>\n\n</div>", "_____no_output_____" ] ], [ [ "data['2013':'2013'].mean().plot(kind='barh')", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code", "code", "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown" ], [ "code", "code", "code", "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "# Importing Spark NLP, Spark NLP for Healthcare and Spark OCR", "_____no_output_____" ] ], [ [ "import sparknlp", "_____no_output_____" ], [ "import sparknlp_jsl", "_____no_output_____" ], [ "import sparkocr", "_____no_output_____" ], [ "sparknlp_jsl.version()", "_____no_output_____" ], [ "sparknlp.version()", "_____no_output_____" ], [ "sparkocr.version()", "_____no_output_____" ] ], [ [ "# Retrieving your license", "_____no_output_____" ] ], [ [ "import os, json\n\nwith open('/license.json', 'r') as f:\n license_keys = json.load(f)\n\n# Defining license key-value pairs as local variables\nlocals().update(license_keys)\n\n# Adding license key-value pairs to environment variables\nos.environ.update(license_keys)", "_____no_output_____" ] ], [ [ "# Add a DNS entry for your Sagemaker instance", "_____no_output_____" ] ], [ [ "!echo \"127.0.0.1 $HOSTNAME\" >> /etc/hosts", "_____no_output_____" ] ], [ [ "# Start your session", "_____no_output_____" ] ], [ [ " spark = sparkocr.start(secret=os.environ['SPARK_OCR_SECRET'], nlp_secret=['SECRET'])", "Spark version: 3.0.2\nSpark NLP version: 3.3.1\nSpark OCR version: 3.8.0\n\n" ] ], [ [ "# Check everything is good and have fun!", "_____no_output_____" ] ], [ [ "spark", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "Demonstrating how to get DonkeyCar Tub files into a PyTorch/fastai DataBlock", "_____no_output_____" ] ], [ [ "from fastai.data.all import *\nfrom fastai.vision.all import *\nfrom fastai.data.transforms import ColReader, Normalize, RandomSplitter\nimport torch\nfrom torch import nn\nfrom torch.nn import functional as F", "_____no_output_____" ], [ "from donkeycar.parts.tub_v2 import Tub\nimport pandas as pd\nfrom pathlib import Path", "_____no_output_____" ], [ "from malpi.dk.train import preprocessFileList, get_data, get_learner, get_autoencoder, train_autoencoder", "_____no_output_____" ], [ "def learn_resnet():\n learn2 = cnn_learner(dls, resnet18, loss_func=MSELossFlat(), metrics=[rmse], cbs=ActivationStats(with_hist=True))\n learn2.fine_tune(5)\n \n learn2.recorder.plot_loss()\n learn2.show_results(figsize=(20,10))", "_____no_output_____" ] ], [ [ "The below code is modified from: https://github.com/cmasenas/fastai_navigation_training/blob/master/fastai_train.ipynb.\n\nTODO: Figure out how to have multiple output heads", "_____no_output_____" ] ], [ [ "def test_one_transform(name, inputs, df_all, batch_tfms, item_tfms, epochs, lr):\n dls = get_data(inputs, df_all=df_all, batch_tfms=batch_tfms, item_tfms=item_tfms)\n callbacks = [CSVLogger(f\"Transform_{name}.csv\", append=True)]\n learn = get_learner(dls)\n #learn.no_logging() #Try this to block logging when doing many training test runs\n learn.fit_one_cycle(epochs, lr, cbs=callbacks)\n #learn.recorder.plot_loss()\n #learn.show_results(figsize=(20,10))", "_____no_output_____" ], [ "# Train multipel times using a list of Transforms, one at a time.\n# Compare mean/stdev of best validation loss (or rmse?) for each Transform\ndf_all = get_dataframe(\"track1_warehouse.txt\")\ntransforms = [None]\ntransforms.extend( [*aug_transforms(do_flip=False, size=128)] )\nfor tfm in transforms:\n name = \"None\" if tfm is None else str(tfm.__class__.__name__)\n print( f\"Transform: {name}\" )\n for i in range(5):\n print( f\" Run {i+1}\" )\n test_one_transform(name, \"track1_warehouse.txt\", df_all, None, 5, 3e-3)", "_____no_output_____" ], [ "def visualize_learner( learn ):\n #dls=nav.dataloaders(df, bs=512)\n preds, tgt = learn.get_preds(dl=[dls.one_batch()])\n\n plt.title(\"Target vs Predicted Steering\", fontsize=18, y=1.0)\n plt.xlabel(\"Target\", fontsize=14, labelpad=15)\n plt.ylabel(\"Predicted\", fontsize=14, labelpad=15)\n plt.plot(tgt.T[0], preds.T[0],'bo')\n plt.plot([-1,1],[-1,1],'r', linewidth = 4)\n plt.show()\n\n plt.title(\"Target vs Predicted Throttle\", fontsize=18, y=1.02)\n plt.xlabel(\"Target\", fontsize=14, labelpad=15)\n plt.ylabel(\"Predicted\", fontsize=14, labelpad=15)\n plt.plot(tgt.T[1], preds.T[1],'bo')\n plt.plot([0,1],[0,1],'r', linewidth = 4)\n plt.show()", "_____no_output_____" ], [ "learn.export()", "_____no_output_____" ], [ "df_all = get_dataframe(\"track1_warehouse.txt\")\ndls = get_data(\"track1_warehouse.txt\", df_all=df_all, batch_tfms=None)", "_____no_output_____" ], [ "learn = get_learner(dls)\nlearn.fit_one_cycle(15, 3e-3)", "_____no_output_____" ], [ "visualize_learner(learn)", "_____no_output_____" ], [ "learn.export('models/track1_v2.pkl')", "_____no_output_____" ], [ "def clear_pyplot_memory():\n plt.clf()\n plt.cla()\n plt.close()\n\ndf_all = get_dataframe(\"track1_warehouse.txt\")\n\ntransforms=[None,\n RandomResizedCrop(128,p=1.0,min_scale=0.5,ratio=(0.9,1.1)),\n RandomErasing(sh=0.2, max_count=6,p=1.0),\n Brightness(max_lighting=0.4, p=1.0),\n Contrast(max_lighting=0.4, p=1.0),\n Saturation(max_lighting=0.4, p=1.0)]\n#dls = get_data(None, df_all, item_tfms=item_tfms, batch_tfms=batch_tfms)\n\nfor tfm in transforms:\n name = \"None\" if tfm is None else str(tfm.__class__.__name__)\n if name == \"RandomResizedCrop\":\n item_tfms = tfm\n batch_tfms = None\n else:\n item_tfms = None\n batch_tfms = tfm\n \n dls = get_data(\"track1_warehouse.txt\",\n df_all=df_all,\n item_tfms=item_tfms, batch_tfms=batch_tfms)\n\n dls.show_batch(unique=True, show=True)\n plt.savefig( f'Transform_{name}.png' )\n#clear_pyplot_memory()", "_____no_output_____" ], [ "learn, dls = train_autoencoder( \"tracks_all.txt\", 5, 3e-3, name=\"ae_test1\", verbose=False )", "_____no_output_____" ], [ "learn.recorder.plot_loss()\nlearn.show_results(figsize=(20,10))\n#plt.savefig(name + '.png')", "_____no_output_____" ], [ "idx = 0", "_____no_output_____" ], [ "idx += 1\nim1 = dls.one_batch()[0]\nim1_out = learn.model.forward(im1)\nshow_image(im1[idx])\nshow_image(im1_out[idx])", "_____no_output_____" ], [ "from fastai.metrics import rmse", "_____no_output_____" ], [ "from typing import List, Callable, Union, Any, TypeVar, Tuple\nTensor = TypeVar('torch.tensor')\n\nfrom abc import abstractmethod\n\nclass BaseVAE(nn.Module):\n\n def __init__(self) -> None:\n super(BaseVAE, self).__init__()\n\n def encode(self, input: Tensor) -> List[Tensor]:\n raise NotImplementedError\n\n def decode(self, input: Tensor) -> Any:\n raise NotImplementedError\n\n def sample(self, batch_size:int, current_device: int, **kwargs) -> Tensor:\n raise NotImplementedError\n\n def generate(self, x: Tensor, **kwargs) -> Tensor:\n raise NotImplementedError\n\n @abstractmethod\n def forward(self, *inputs: Tensor) -> Tensor:\n pass\n\n @abstractmethod\n def loss_function(self, *inputs: Any, **kwargs) -> Tensor:\n pass", "_____no_output_____" ], [ "class VanillaVAE(BaseVAE):\n\n\n def __init__(self,\n in_channels: int,\n latent_dim: int,\n hidden_dims: List = None,\n **kwargs) -> None:\n super(VanillaVAE, self).__init__()\n\n self.latent_dim = latent_dim\n self.kld_weight = 0.00025 # TODO calculate based on: #al_img.shape[0]/ self.num_train_imgs\n modules = []\n if hidden_dims is None:\n hidden_dims = [32, 64, 128, 256, 512]\n\n # Build Encoder\n for h_dim in hidden_dims:\n modules.append(\n nn.Sequential(\n nn.Conv2d(in_channels, out_channels=h_dim,\n kernel_size= 3, stride= 2, padding = 1),\n nn.BatchNorm2d(h_dim),\n nn.LeakyReLU())\n )\n in_channels = h_dim\n\n self.encoder = nn.Sequential(*modules)\n self.fc_mu = nn.Linear(hidden_dims[-1]*4, latent_dim)\n self.fc_var = nn.Linear(hidden_dims[-1]*4, latent_dim)\n\n\n # Build Decoder\n modules = []\n\n self.decoder_input = nn.Linear(latent_dim, hidden_dims[-1] * 4)\n\n hidden_dims.reverse()\n\n for i in range(len(hidden_dims) - 1):\n modules.append(\n nn.Sequential(\n nn.ConvTranspose2d(hidden_dims[i],\n hidden_dims[i + 1],\n kernel_size=3,\n stride = 2,\n padding=1,\n output_padding=1),\n nn.BatchNorm2d(hidden_dims[i + 1]),\n nn.LeakyReLU())\n )\n\n\n\n self.decoder = nn.Sequential(*modules)\n\n self.final_layer = nn.Sequential(\n nn.ConvTranspose2d(hidden_dims[-1],\n hidden_dims[-1],\n kernel_size=3,\n stride=2,\n padding=1,\n output_padding=1),\n nn.BatchNorm2d(hidden_dims[-1]),\n nn.LeakyReLU(),\n nn.Conv2d(hidden_dims[-1], out_channels= 3,\n kernel_size= 3, padding= 1),\n nn.Tanh())\n\n def encode(self, input: Tensor) -> List[Tensor]:\n \"\"\"\n Encodes the input by passing through the encoder network\n and returns the latent codes.\n :param input: (Tensor) Input tensor to encoder [N x C x H x W]\n :return: (Tensor) List of latent codes\n \"\"\"\n result = self.encoder(input)\n result = torch.flatten(result, start_dim=1)\n\n # Split the result into mu and var components\n # of the latent Gaussian distribution\n mu = self.fc_mu(result)\n log_var = self.fc_var(result)\n\n return [mu, log_var]\n\n def decode(self, z: Tensor) -> Tensor:\n \"\"\"\n Maps the given latent codes\n onto the image space.\n :param z: (Tensor) [B x D]\n :return: (Tensor) [B x C x H x W]\n \"\"\"\n result = self.decoder_input(z)\n result = result.view(-1, 512, 2, 2)\n result = self.decoder(result)\n result = self.final_layer(result)\n return result\n\n def reparameterize(self, mu: Tensor, logvar: Tensor) -> Tensor:\n \"\"\"\n Reparameterization trick to sample from N(mu, var) from\n N(0,1).\n :param mu: (Tensor) Mean of the latent Gaussian [B x D]\n :param logvar: (Tensor) Standard deviation of the latent Gaussian [B x D]\n :return: (Tensor) [B x D]\n \"\"\"\n std = torch.exp(0.5 * logvar)\n eps = torch.randn_like(std)\n return eps * std + mu\n\n def forward(self, input: Tensor, **kwargs) -> List[Tensor]:\n mu, log_var = self.encode(input)\n z = self.reparameterize(mu, log_var)\n return [self.decode(z), input, mu, log_var]\n\n def loss_function(self,\n *args,\n **kwargs) -> dict:\n \"\"\"\n Computes the VAE loss function.\n KL(N(\\mu, \\sigma), N(0, 1)) = \\log \\frac{1}{\\sigma} + \\frac{\\sigma^2 + \\mu^2}{2} - \\frac{1}{2}\n :param args:\n :param kwargs:\n :return:\n \"\"\"\n #print( f\"loss_function: {len(args[0])} {type(args[0][0])} {args[1].shape}\" )\n recons = args[0][0]\n input = args[1]\n mu = args[0][2]\n log_var = args[0][3]\n\n kld_weight = self.kld_weight # kwargs['M_N'] # Account for the minibatch samples from the dataset\n recons_loss =F.mse_loss(recons, input)\n\n\n kld_loss = torch.mean(-0.5 * torch.sum(1 + log_var - mu ** 2 - log_var.exp(), dim = 1), dim = 0)\n\n loss = recons_loss + kld_weight * kld_loss\n return loss\n #return {'loss': loss, 'Reconstruction_Loss':recons_loss.detach(), 'KLD':-kld_loss.detach()}\n\n def sample(self,\n num_samples:int,\n current_device: int, **kwargs) -> Tensor:\n \"\"\"\n Samples from the latent space and return the corresponding\n image space map.\n :param num_samples: (Int) Number of samples\n :param current_device: (Int) Device to run the model\n :return: (Tensor)\n \"\"\"\n z = torch.randn(num_samples,\n self.latent_dim)\n\n z = z.to(current_device)\n\n samples = self.decode(z)\n return samples\n\n def generate(self, x: Tensor, **kwargs) -> Tensor:\n \"\"\"\n Given an input image x, returns the reconstructed image\n :param x: (Tensor) [B x C x H x W]\n :return: (Tensor) [B x C x H x W]\n \"\"\"\n\n return self.forward(x)[0]\n", "_____no_output_____" ], [ "input_file=\"track1_warehouse.txt\"\nitem_tfms = [Resize(64,method=\"squish\")]\ndls = get_data(input_file, item_tfms=item_tfms, verbose=False, autoencoder=True)", "_____no_output_____" ], [ "vae = VanillaVAE(3, 64)\nlearn = Learner(dls, vae, loss_func=vae.loss_function)", "_____no_output_____" ], [ "learn.fit_one_cycle(5, 3e-3)", "_____no_output_____" ], [ "vae", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import matplotlib.pyplot as plt ### fig:class_world1", "_____no_output_____" ], [ "class World: \n def __init__(self):\n self.objects = [] # ここにロボットなどのオブジェクトを登録\n \n def append(self,obj): # オブジェクトを登録するための関数\n self.objects.append(obj)\n \n def draw(self):\n fig = plt.figure(figsize=(8,8)) # 8x8 inchの図を準備\n ax = fig.add_subplot(111) # サブプロットを準備\n ax.set_aspect('equal') # 縦横比を座標の値と一致させる\n ax.set_xlim(-5,5) # X軸を-5m x 5mの範囲で描画\n ax.set_ylim(-5,5) # Y軸も同様に\n ax.set_xlabel(\"X\",fontsize=20) # X軸にラベルを表示\n ax.set_ylabel(\"Y\",fontsize=20) # 同じくY軸に\n \n for obj in self.objects: obj.draw(ax) # appendした物体を次々に描画\n \n plt.show()", "_____no_output_____" ], [ "world = World() ### fig:class_world3\nworld.draw()", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Redis列表实现一次pop 弹出多条数据\n\n\n", "_____no_output_____" ] ], [ [ "# 连接 Redis\n\nimport redis\nclient = redis.Redis(host='122.51.39.219', port=6379, password='leftright123')\n\n# 注意：\n# 这个 Redis 环境仅作为练习之用，每小时会清空一次，请勿存放重要数据。", "_____no_output_____" ], [ "# 准备数据\n\nclient.lpush('test_batch_pop', *list(range(10000)))", "_____no_output_____" ], [ "# 一条一条读取，非常耗时\nimport time\n\n\nstart = time.time()\nwhile True:\n data = client.lpop('test_batch_pop')\n if not data:\n break\nend = time.time()\n\ndelta = end - start\nprint(f'循环读取10000条数据，使用 lpop 耗时：{delta}')", "循环读取10000条数据，使用 lpop 耗时：112.04084920883179\n" ] ], [ [ "## 为什么使用`lpop`读取10000条数据这么慢？\n\n因为`lpop`每次只弹出1条数据，每次弹出数据都要连接 Redis 。大量时间浪费在了网络传输上面。\n\n## 如何实现批量弹出多条数据，并在同一次网络请求中返回？\n\n先使用 `lrange` 获取数据，再使用`ltrim`删除被获取的数据。", "_____no_output_____" ] ], [ [ "# 复习一下 lrange 的用法\n\ndatas = client.lrange('test_batch_pop', 0, 9) # 读取前10条数据\ndatas", "_____no_output_____" ], [ "# 学习一下 ltrim 的用法\n\nclient.ltrim('test_batch_pop', 10, -1) # 删除前10条数据", "_____no_output_____" ], [ "# 验证一下数据是否被成功删除\n\nlength = client.llen('test_batch_pop')\nprint(f'现在列表里面还剩{length}条数据')\ndatas = client.lrange('test_batch_pop', 0, 9) # 读取前10条数据\ndatas", "现在列表里面还剩9990条数据\n" ], [ "# 一种看起来正确的做法\n\ndef batch_pop_fake(key, n):\n datas = client.lrange(key, 0, n - 1)\n client.ltrim(key, n, -1)\n return datas\n\nbatch_pop_fake('test_batch_pop', 10)", "_____no_output_____" ], [ "client.lrange('test_batch_pop', 0, 9)", "_____no_output_____" ] ], [ [ "## 这种写法用什么问题\n\n在多个进程同时使用 batch_pop_fake 函数的时候，由于执行 lrange 与 ltrim 是在两条语句中，因此实际上会分成2个网络请求。那么当 A 进程\n刚刚执行完lrange，还没有来得及执行 ltrim 时，B 进程刚好过来执行 lrange，那么 AB 两个进程就会获得相同的数据。\n\n等 B 进程获取完成数据以后，A 进程的 ltrim 刚刚抵达，此时Redis 会删除前 n 条数据，然后 B 进程的 ltrim 也到了，再删除前 n 条数据。那么最终导致的结果就是，AB 两个进程同时拿到前 n 条数据，但是却有2n 条数据被删除。", "_____no_output_____" ], [ "## 使用 pipeline 打包多个命令到一个请求中\n\npipeline 的使用方法如下：\n\n```python\nimport redis\n\nclient = redis.Redis()\npipe = client.pipeline()\npipe.lrange('key', 0, n - 1)\npipe.ltrim('key', n, -1)\nresult = pipe.execute()\n```\n\npipe.execute()返回一个列表，这个列表每一项按顺序对应每一个命令的执行结果。在上面的例子中，result 是一个有两项的列表，第一项对应 lrange 的返回结果，第二项为 True，表示 ltrim 执行成功。", "_____no_output_____" ] ], [ [ "# 真正可用的批量弹出数据函数\n\ndef batch_pop_real(key, n):\n pipe = client.pipeline()\n pipe.lrange(key, 0, n - 1)\n pipe.ltrim(key, n, -1)\n result = pipe.execute()\n return result[0]", "_____no_output_____" ], [ "# 清空列表并重新添加10000条数据\nclient.delete('test_batch_pop')\nclient.lpush('test_batch_pop', *list(range(10000)))", "_____no_output_____" ], [ "start = time.time()\nwhile True:\n datas = batch_pop_real('test_batch_pop', 1000)\n if not datas:\n break\n for data in datas:\n pass\nend = time.time()\nprint(f'批量弹出10000条数据，耗时：{end - start}')", "批量弹出10000条数据，耗时：0.18534111976623535\n" ], [ "client.llen('test_batch_pop')", "_____no_output_____" ] ], [ [ "\n\n", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import pandas as pd", "_____no_output_____" ], [ "df = pd.read_excel('penjualan.xlsx', usecols=[1, 2, 3], skiprows=[0])", "_____no_output_____" ], [ "df.head()", "_____no_output_____" ], [ "df.dtypes", "_____no_output_____" ], [ "df.loc[0, 'JUMLAH'] = 47000", "_____no_output_____" ], [ "df", "_____no_output_____" ], [ "rekap_per_bulan = df.groupby(df.TGL.dt.month).sum()", "_____no_output_____" ], [ "rekap_per_bulan", "_____no_output_____" ], [ "df['JUMLAH'].sum()", "_____no_output_____" ], [ "rekap_per_bulan.sum()", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "# importamos las librerías necesarias\n%matplotlib inline\nimport random\nimport tsfresh\nimport os\nimport math\nfrom scipy import stats\nfrom scipy.spatial.distance import pdist\nfrom math import sqrt, log, floor\nfrom fastdtw import fastdtw\nimport ipywidgets as widgets\nimport matplotlib.pyplot as plt\nimport matplotlib.cm as cm\nimport numpy as np\nimport pandas as pd\nimport seaborn as sns\nfrom statistics import mean\nfrom scipy.spatial.distance import euclidean\nimport scipy.cluster.hierarchy as hac\nfrom scipy.cluster.hierarchy import fcluster\nfrom sklearn.metrics.pairwise import pairwise_distances\nfrom sklearn.decomposition import PCA\nfrom sklearn.cluster import KMeans, AgglomerativeClustering, DBSCAN\nfrom sklearn.manifold import TSNE\nfrom sklearn.metrics import normalized_mutual_info_score, adjusted_rand_score, silhouette_score, silhouette_samples\nfrom sklearn.metrics import mean_squared_error\nfrom scipy.spatial import distance\n\nsns.set(style='white')\n# \"fix\" the randomness for reproducibility\nrandom.seed(42)", "_____no_output_____" ], [ "!pip install tsfresh", "Collecting tsfresh\n Downloading tsfresh-0.17.0-py2.py3-none-any.whl (91 kB)\nRequirement already satisfied: statsmodels>=0.9.0 in c:\\users\\yoda\\anaconda3\\envs\\tensorflow2\\lib\\site-packages (from tsfresh) (0.12.1)\nRequirement already satisfied: patsy>=0.4.1 in c:\\users\\yoda\\anaconda3\\envs\\tensorflow2\\lib\\site-packages (from tsfresh) (0.5.1)\nRequirement already satisfied: scipy>=1.2.0 in c:\\users\\yoda\\anaconda3\\envs\\tensorflow2\\lib\\site-packages (from tsfresh) (1.4.1)\nRequirement already satisfied: pandas>=0.25.0 in c:\\users\\yoda\\anaconda3\\envs\\tensorflow2\\lib\\site-packages (from tsfresh) (1.0.5)\nCollecting dask[dataframe]>=2.9.0\n Downloading dask-2.30.0-py3-none-any.whl (848 kB)\nRequirement already satisfied: requests>=2.9.1 in c:\\users\\yoda\\anaconda3\\envs\\tensorflow2\\lib\\site-packages (from tsfresh) (2.23.0)\nRequirement already satisfied: numpy>=1.15.1 in c:\\users\\yoda\\anaconda3\\envs\\tensorflow2\\lib\\site-packages (from tsfresh) (1.18.5)\nCollecting tqdm>=4.10.0\n Downloading tqdm-4.51.0-py2.py3-none-any.whl (70 kB)\nCollecting distributed>=2.11.0\n Downloading distributed-2.30.1-py3-none-any.whl (656 kB)\nRequirement already satisfied: scikit-learn>=0.19.2 in c:\\users\\yoda\\anaconda3\\envs\\tensorflow2\\lib\\site-packages (from tsfresh) (0.22.1)\nRequirement already satisfied: six in c:\\users\\yoda\\anaconda3\\envs\\tensorflow2\\lib\\site-packages (from patsy>=0.4.1->tsfresh) (1.15.0)\nRequirement already satisfied: python-dateutil>=2.6.1 in c:\\users\\yoda\\anaconda3\\envs\\tensorflow2\\lib\\site-packages (from pandas>=0.25.0->tsfresh) (2.8.1)\nRequirement already satisfied: pytz>=2017.2 in c:\\users\\yoda\\anaconda3\\envs\\tensorflow2\\lib\\site-packages (from pandas>=0.25.0->tsfresh) (2020.1)\nRequirement already satisfied: pyyaml in c:\\users\\yoda\\anaconda3\\envs\\tensorflow2\\lib\\site-packages (from dask[dataframe]>=2.9.0->tsfresh) (5.3.1)\nCollecting toolz>=0.8.2; extra == \"dataframe\"\n Downloading toolz-0.11.1-py3-none-any.whl (55 kB)\nCollecting fsspec>=0.6.0; extra == \"dataframe\"\n Downloading fsspec-0.8.4-py3-none-any.whl (91 kB)\nCollecting partd>=0.3.10; extra == \"dataframe\"\n Downloading partd-1.1.0-py3-none-any.whl (19 kB)\nRequirement already satisfied: chardet<4,>=3.0.2 in c:\\users\\yoda\\anaconda3\\envs\\tensorflow2\\lib\\site-packages (from requests>=2.9.1->tsfresh) (3.0.4)\nRequirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in c:\\users\\yoda\\anaconda3\\envs\\tensorflow2\\lib\\site-packages (from requests>=2.9.1->tsfresh) (1.25.9)\nRequirement already satisfied: idna<3,>=2.5 in c:\\users\\yoda\\anaconda3\\envs\\tensorflow2\\lib\\site-packages (from requests>=2.9.1->tsfresh) (2.9)\nRequirement already satisfied: certifi>=2017.4.17 in c:\\users\\yoda\\anaconda3\\envs\\tensorflow2\\lib\\site-packages (from requests>=2.9.1->tsfresh) (2020.6.20)\nCollecting zict>=0.1.3\n Downloading zict-2.0.0-py3-none-any.whl (10 kB)\nCollecting cloudpickle>=1.5.0\n Downloading cloudpickle-1.6.0-py3-none-any.whl (23 kB)\nCollecting psutil>=5.0\n Downloading psutil-5.7.3-cp37-cp37m-win_amd64.whl (243 kB)\nRequirement already satisfied: setuptools in c:\\users\\yoda\\anaconda3\\envs\\tensorflow2\\lib\\site-packages (from distributed>=2.11.0->tsfresh) (47.3.0.post20200616)\nCollecting sortedcontainers!=2.0.0,!=2.0.1\n Downloading sortedcontainers-2.2.2-py2.py3-none-any.whl (29 kB)\nCollecting msgpack>=0.6.0\n Downloading msgpack-1.0.0-cp37-cp37m-win_amd64.whl (72 kB)\nRequirement already satisfied: tornado>=5; python_version < \"3.8\" in c:\\users\\yoda\\anaconda3\\envs\\tensorflow2\\lib\\site-packages (from distributed>=2.11.0->tsfresh) (6.0.4)\nCollecting tblib>=1.6.0\n Downloading tblib-1.7.0-py2.py3-none-any.whl (12 kB)\nRequirement already satisfied: click>=6.6 in c:\\users\\yoda\\anaconda3\\envs\\tensorflow2\\lib\\site-packages (from distributed>=2.11.0->tsfresh) (7.1.2)\nRequirement already satisfied: joblib>=0.11 in c:\\users\\yoda\\anaconda3\\envs\\tensorflow2\\lib\\site-packages (from scikit-learn>=0.19.2->tsfresh) (0.15.1)\nCollecting locket\n Downloading locket-0.2.0.tar.gz (3.5 kB)\nCollecting heapdict\n Downloading HeapDict-1.0.1-py3-none-any.whl (3.9 kB)\nBuilding wheels for collected packages: locket\n Building wheel for locket (setup.py): started\n Building wheel for locket (setup.py): finished with status 'done'\n Created wheel for locket: filename=locket-0.2.0-py3-none-any.whl size=4045 sha256=db38e5a4aa8529fa1f790ebe2bcc94b05ffeeccb4159a26e2933fbd080a63a4a\n Stored in directory: c:\\users\\yoda\\appdata\\local\\pip\\cache\\wheels\\7d\\21\\ce\\56f01c644a11bde5d09ecae16d9b5e9d7e988187624fd28fec\nSuccessfully built locket\nInstalling collected packages: toolz, fsspec, locket, partd, dask, tqdm, heapdict, zict, cloudpickle, psutil, sortedcontainers, msgpack, tblib, distributed, tsfresh\nSuccessfully installed cloudpickle-1.6.0 dask-2.30.0 distributed-2.30.1 fsspec-0.8.4 heapdict-1.0.1 locket-0.2.0 msgpack-1.0.0 partd-1.1.0 psutil-5.7.3 sortedcontainers-2.2.2 tblib-1.7.0 toolz-0.11.1 tqdm-4.51.0 tsfresh-0.17.0 zict-2.0.0\n" ] ], [ [ "### Dataset", "_____no_output_____" ], [ "Los datos son series temporales (casos semanales de Dengue) de distintos distritos de Paraguay", "_____no_output_____" ] ], [ [ "path = \"./data/Notificaciones/\"\nfilename_read = os.path.join(path,\"normalizado.csv\")\nnotificaciones = pd.read_csv(filename_read,delimiter=\",\",engine='python')\nnotificaciones.shape", "_____no_output_____" ], [ "listaMunicp = notificaciones['distrito_nombre'].tolist()\nlistaMunicp = list(dict.fromkeys(listaMunicp))\nprint('Son ', len(listaMunicp), ' distritos')\nlistaMunicp.sort()\nprint(listaMunicp)", "Son 217 distritos\n['1RO DE MARZO', '25 DE DICIEMBRE', '3 DE FEBRERO', 'ABAI', 'ACAHAY', 'ALBERDI', 'ALTO VERA', 'ALTOS', 'ANTEQUERA', 'AREGUA', 'ARROYOS Y ESTEROS', 'ASUNCION', 'ATYRA', 'AYOLAS', 'AZOTEY', 'BAHIA NEGRA', 'BELEN', 'BELLA VISTA', 'BENJAMIN ACEVAL', 'BORJA', 'BUENA VISTA', 'CAACUPE', 'CAAGUAZU', 'CAAZAPA', 'CABALLERO ALVAREZ', 'CAMBYRETA', 'CAPIATA', 'CAPIIBARY', 'CAPITAN BADO', 'CAPITAN MEZA', 'CAPITAN MIRANDA', 'CARAGUATAY', 'CARAPEGUA', 'CARAYAO', 'CARLOS ANTONIO LOPEZ', 'CARMELO PERALTA', 'CARMEN DEL PARANA', 'CECILIO BAEZ', 'CERRITO', 'CHACO', 'CHORE', 'COLONIA FRAM', 'COLONIA INDEPENDENCIA', 'CONCEPCION', 'CORONEL BOGADO', 'CORONEL MARTINEZ', 'CORONEL OVIEDO', 'CORPUS CHRISTI', 'CURUGUATY', 'DESMOCHADOS', 'DR BOTRELL', 'DR. JUAN MANUEL FRUTOS', 'EDELIRA', 'EMBOSCADA', 'ENCARNACION', 'ESCOBAR', 'EUGENIO A GARAY', 'EUSEBIO AYALA', 'FASSARDI', 'FELIX PEREZ CARDOZO', 'FERNANDO DE LA MORA', 'FILADELFIA', 'FUERTE OLIMPO', 'GENERAL AQUINO', 'GENERAL ARTIGAS', 'GENERAL BERNARDINO CABALLERO', 'GENERAL BRUGUEZ', 'GENERAL DELGADO', 'GENERAL DIAZ', 'GENERAL MORINIGO', 'GENERAL RESQUIN', 'GUARAMBARE', 'GUAYAIBI', 'GUAZUCUA', 'HERNANDARIAS', 'HOHENAU', 'HORQUETA', 'HUMAITA', 'ISLA PUCU', 'ISLA UMBU', 'ITA', 'ITACURUBI DE LA CORDILLERA', 'ITACURUBI DEL ROSARIO', 'ITAKYRY', 'ITANARA', 'ITAPE', 'ITAPUA POTY', 'ITAUGUA', 'ITURBE', 'J A SALDIVAR', 'JESUS', 'JOSE DOMINGO OCAMPOS', 'JUAN DE MENA', 'JUAN E. OLEARY', 'JUAN EULOGIO ESTIGARRIBIA', 'JUAN LEON MALLORQUIN', 'KATUETE', 'LA PALOMA', 'LA PASTORA', 'LA VICTORIA', 'LAMBARE', 'LAURELES', 'LEANDRO OVIEDO', 'LIMA', 'LIMOY PUEBLO', 'LIMPIO', 'LOMA GRANDE', 'LOMA PLATA', 'LORETO', 'LUQUE', 'MACIEL', 'MARIANO ROQUE ALONSO', 'MAURICIO JOSE TROCHE', 'MBARACAYU', 'MBOCAYATY', 'MBOCAYATY DEL YHAGUY', 'MCAL. ESTIGARRIBIA', 'MCAL. FRANCISCO SOLANO LOPEZ', 'MINGA GUAZU', 'MINGA PORA', 'MOISES BERTONI', 'NANAWA', 'NARANJAL', 'NATALICIO TALAVERA', 'NATALIO', 'NUEVA ALBORADA', 'NUEVA COLOMBIA', 'NUEVA ESPERANZA', 'NUEVA GERMANIA', 'NUEVA ITALIA', 'NUEVA LONDRES', 'OBLIGADO', 'PARAGUARI', 'PASO YOBAI', 'PEDRO JUAN CABALLERO', 'PILAR', 'PIRAPO', 'PIRAYU', 'PIRIBEBUY', 'POZO COLORADO', 'PUERTO FALCON', 'PUERTO PINASCO', 'QUIINDY', 'R I 3 CORRALES', 'RAUL ARSENIO OVIEDO', 'REPATRIACION', 'ROQUE GONZALEZ DE SANTA CRUZ', 'SALTO DEL GUAIRA', 'SAN ALBERTO', 'SAN ANTONIO', 'SAN BERNARDINO', 'SAN CARLOS', 'SAN COSME Y DAMIAN', 'SAN ESTANISLAO', 'SAN IGNACIO', 'SAN JOAQUIN', 'SAN JOSE DE LOS ARROYOS', 'SAN JOSE OBRERO', 'SAN JUAN BAUTISTA', 'SAN JUAN DEL PARANA', 'SAN JUAN NEPOMUCENO', 'SAN LAZARO', 'SAN LORENZO', 'SAN MIGUEL', 'SAN PATRICIO', 'SAN PEDRO', 'SAN PEDRO DEL PARANA', 'SAN PEDRO DEL YCUAMANDIYU', 'SAN RAFAEL DEL PARANA', 'SAN ROQUE GONZALEZ DE SANTACRUZ', 'SAN SALVADOR', 'SANTA ELENA', 'SANTA MARIA', 'SANTA RITA', 'SANTA ROSA', 'SANTA ROSA DEL AGUARAY', 'SANTA ROSA DEL MBUTUY', 'SANTA ROSA DEL MONDAY', 'SANTIAGO', 'SAPUCAI', 'SIMON BOLIVAR', 'TACUARAS', 'TACUATI', 'TAVAI', 'TAVAPY', 'TEBICUARY', 'TEBICUARYMI', 'TEMBIAPORA', 'TOBATI', 'TOMAS ROMERO PEREIRA', 'TRINIDAD', 'UNION', 'VALENZUELA', 'VAQUERIA', 'VILLA DEL ROSARIO', 'VILLA ELISA', 'VILLA HAYES', 'VILLA OLIVA', 'VILLALBIN', 'VILLARRICA', 'VILLETA', 'YAGUARON', 'YATAITY', 'YATAITY DEL NORTE', 'YATYTAY', 'YBY YAU', 'YBYRAROVANA', 'YBYTYMI', 'YEGROS', 'YGATIMI', 'YGUAZU', 'YHU', 'YPACARAI', 'YPANE', 'YPEJHU', 'YUTY', 'ZANJA PYTA']\n" ] ], [ [ "A continuación tomamos las series temporales que leímos y vemos como quedan", "_____no_output_____" ] ], [ [ "timeSeries = pd.DataFrame()\nfor muni in listaMunicp:\n municipio=notificaciones['distrito_nombre']==muni\n notif_x_municp=notificaciones[municipio]\n notif_x_municp = notif_x_municp.reset_index(drop=True)\n notif_x_municp = notif_x_municp['incidencia']\n notif_x_municp = notif_x_municp.replace('nan', np.nan).fillna(0.000001)\n notif_x_municp = notif_x_municp.replace([np.inf, -np.inf], np.nan).fillna(0.000001)\n timeSeries = timeSeries.append(notif_x_municp)\n ax = sns.tsplot(ax=None, data=notif_x_municp.values, err_style=\"unit_traces\")\nplt.show()", "/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/seaborn/timeseries.py:183: UserWarning: The `tsplot` function is deprecated and will be removed in a future release. Please update your code to use the new `lineplot` function.\n warnings.warn(msg, UserWarning)\n" ], [ "#timeseries shape\nn=217", "_____no_output_____" ], [ "timeSeries.shape", "_____no_output_____" ], [ "timeSeries.describe()", "_____no_output_____" ] ], [ [ "### Análisis de grupos (Clustering)", "_____no_output_____" ], [ "El Clustering o la clusterización es un proceso importante dentro del Machine learning. Este proceso desarrolla una acción fundamental que le permite a los algoritmos de aprendizaje automatizado entrenar y conocer de forma adecuada los datos con los que desarrollan sus actividades. Tiene como finalidad principal lograr el agrupamiento de conjuntos de objetos no etiquetados, para lograr construir subconjuntos de datos conocidos como Clusters. Cada cluster dentro de un grafo está formado por una colección de objetos o datos que a términos de análisis resultan similares entre si, pero que poseen elementos diferenciales con respecto a otros objetos pertenecientes al conjunto de datos y que pueden conformar un cluster independiente.", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "Aunque los datos no necesariamente son tan fáciles de agrupar", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "### Métricas de similitud", "_____no_output_____" ], [ " Para medir lo similares ( o disimilares) que son los individuos existe una enorme cantidad de índices de similaridad y de disimilaridad o divergencia. Todos ellos tienen propiedades y utilidades distintas y habrá que ser consciente de ellas para su correcta aplicación al caso que nos ocupe.\nLa mayor parte de estos índices serán o bien, indicadores basados en la distancia (considerando a los individuos como vectores en el espacio de las variables) (en este sentido un elevado valor de la distancia entre dos individuos nos indicará un alto grado de disimilaridad entre ellos); o bien, indicadores basados en coeficientes de correlación ; o bien basados en tablas de datos de posesión o no de una serie de atributos. \nA continuación mostramos las funciones de: \n* Distancia Euclidiana\n* Error cuadrático medio\n* Fast Dynamic Time Warping\n* Correlación de Pearson y\n* Correlación de Spearman.\n\nExisten muchas otras métricas y depende de la naturaleza de cada problema decidir cuál usar. Por ejemplo, *Fast Dymanic Time Warping* es una medida de similitud diseña especialmente para series temporales.\n", "_____no_output_____" ] ], [ [ "#Euclidean\ndef euclidean(x, y):\n r=np.linalg.norm(x-y)\n if math.isnan(r):\n r=1\n #print(r)\n return r", "_____no_output_____" ], [ "#RMSE\ndef rmse(x, y):\n r=sqrt(mean_squared_error(x,y))\n if math.isnan(r):\n r=1\n #print(r)\n return r", "_____no_output_____" ], [ "#Fast Dynamic time warping\ndef fast_DTW(x, y):\n r, _ = fastdtw(x, y, dist=euclidean)\n if math.isnan(r):\n r=1\n #print(r)\n return r", "_____no_output_____" ], [ "#Correlation\ndef corr(x, y):\n r=np.dot(x-mean(x),y-mean(y))/((np.linalg.norm(x-mean(x)))*(np.linalg.norm(y-mean(y))))\n if math.isnan(r):\n r=0\n #print(r)\n return 1 - r", "_____no_output_____" ], [ "#Spearman\ndef scorr(x, y):\n r = stats.spearmanr(x, y)[0]\n if math.isnan(r):\n r=0\n #print(r)\n return 1 - r", "_____no_output_____" ], [ "# compute distances using LCSS\n\n# function for LCSS computation\n# based on implementation from\n# https://rosettacode.org/wiki/Longest_common_subsequence\ndef lcs(a, b): \n lengths = [[0 for j in range(len(b)+1)] for i in range(len(a)+1)]\n # row 0 and column 0 are initialized to 0 already\n for i, x in enumerate(a):\n for j, y in enumerate(b):\n if x == y:\n lengths[i+1][j+1] = lengths[i][j] + 1\n else:\n lengths[i+1][j+1] = max(lengths[i+1][j], lengths[i][j+1])\n x, y = len(a), len(b)\n result = lengths[x][y]\n return result\n\ndef discretise(x):\n return int(x * 10)\n\ndef multidim_lcs(a, b):\n a = a.applymap(discretise)\n b = b.applymap(discretise)\n rows, dims = a.shape\n lcss = [lcs(a[i+2], b[i+2]) for i in range(dims)]\n return 1 - sum(lcss) / (rows * dims)", "_____no_output_____" ], [ "#Distancias para kmeans\n#Euclidean\neuclidean_dist = np.zeros((n,n))\nfor i in range(0,n):\n #print(\"i\",i)\n for j in range(0,n):\n # print(\"j\",j)\n euclidean_dist[i,j] = euclidean(timeSeries.iloc[i].values.flatten(), timeSeries.iloc[j].values.flatten())\n#RMSE\nrmse_dist = np.zeros((n,n))\nfor i in range(0,n):\n #print(\"i\",i)\n for j in range(0,n):\n # print(\"j\",j)\n rmse_dist[i,j] = rmse(timeSeries.iloc[i].values.flatten(), timeSeries.iloc[j].values.flatten())\n#Corr\ncorr_dist = np.zeros((n,n))\nfor i in range(0,n):\n #print(\"i\",i)\n for j in range(0,n):\n # print(\"j\",j)\n corr_dist[i,j] = corr(timeSeries.iloc[i].values.flatten(), timeSeries.iloc[j].values.flatten())\n#scorr\nscorr_dist = np.zeros((n,n))\nfor i in range(0,n):\n #print(\"i\",i)\n for j in range(0,n):\n # print(\"j\",j)\n scorr_dist[i,j] = scorr(timeSeries.iloc[i].values.flatten(), timeSeries.iloc[j].values.flatten())\n#DTW\ndtw_dist = np.zeros((n,n))\nfor i in range(0,n):\n #print(\"i\",i)\n for j in range(0,n):\n # print(\"j\",j)\n dtw_dist[i,j] = fast_DTW(timeSeries.iloc[i].values.flatten(), timeSeries.iloc[j].values.flatten())", "_____no_output_____" ] ], [ [ "### Determinar el número de clusters a formar", "_____no_output_____" ], [ "La mayoría de las técnicas de clustering necesitan como *input* el número de clusters a formar, para eso lo que se hace es hacer una prueba con diferentes números de cluster y nos quedamos con el que dió menor error en general. Para medir ese error utilizamos **Silhouette score**.", "_____no_output_____" ], [ "El **Silhoutte score** se puede utilizar para estudiar la distancia de separación entre los clusters resultantes, especialmente si no hay conocimiento previo de cuáles son los verdaderos grupos para cada objeto, que es el caso más común en aplicaciones reales.", "_____no_output_____" ], [ "El Silhouette score $s(i)$ se calcula:\n\\begin{equation}\ns(i)=\\dfrac{b(i)-a(i)}{max(b(i),a(i))} \n\\end{equation}", "_____no_output_____" ], [ "Definamos $a (i)$ como la distancia media del punto $(i)$ a todos los demás puntos del grupo que se le asignó ($A$). Podemos interpretar $a (i)$ como qué tan bien se asigna el punto al grupo. Cuanto menor sea el valor, mejor será la asignación.\nDe manera similar, definamos $b (i)$ como la distancia media del punto $(i)$ a otros puntos de su grupo vecino más cercano ($B$). El grupo ($B$) es el grupo al que no se asigna el punto $(i)$ pero su distancia es la más cercana entre todos los demás grupos. $ s (i) $ se encuentra en el rango de [-1,1].", "_____no_output_____" ] ], [ [ "from yellowbrick.cluster import KElbowVisualizer\nmodel = AgglomerativeClustering()\nvisualizer = KElbowVisualizer(model, k=(3,20),metric='distortion', timings=False)\n\nvisualizer.fit(rmse_dist) # Fit the data to the visualizer\nvisualizer.show() # Finalize and render the figure", "/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/sklearn/utils/deprecation.py:144: FutureWarning: The sklearn.metrics.classification module is deprecated in version 0.22 and will be removed in version 0.24. The corresponding classes / functions should instead be imported from sklearn.metrics. Anything that cannot be imported from sklearn.metrics is now part of the private API.\n warnings.warn(message, FutureWarning)\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n" ] ], [ [ "Así tenemos que son 9 los grupos que formaremos", "_____no_output_____" ] ], [ [ "k=9", "_____no_output_____" ] ], [ [ "## Técnicas de clustering", "_____no_output_____" ], [ "### K-means", "_____no_output_____" ], [ " El objetivo de este algoritmo es el de encontrar “K” grupos (clusters) entre los datos crudos. El algoritmo trabaja iterativamente para asignar a cada “punto” (las filas de nuestro conjunto de entrada forman una coordenada) uno de los “K” grupos basado en sus características. Son agrupados en base a la similitud de sus features (las columnas). Como resultado de ejecutar el algoritmo tendremos:\n\n* Los “centroids” de cada grupo que serán unas “coordenadas” de cada uno de los K conjuntos que se utilizarán para poder etiquetar nuevas muestras.\n* Etiquetas para el conjunto de datos de entrenamiento. Cada etiqueta perteneciente a uno de los K grupos formados.\n\nLos grupos se van definiendo de manera “orgánica”, es decir que se va ajustando su posición en cada iteración del proceso, hasta que converge el algoritmo. Una vez hallados los centroids deberemos analizarlos para ver cuales son sus características únicas, frente a la de los otros grupos. ", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "En la figura de arriba vemos como los datos se agrupan según el *centroid* que está representado por una estrella. El algortimo inicializa los centroides aleatoriamente y va ajustandolo en cada iteracción, los puntos que están más cerca del *centroid* son los que pertenecen al mismo grupo. ", "_____no_output_____" ], [ "### Clustering jerárquico", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "El algortimo de clúster jerárquico agrupa los datos basándose en la distancia entre cada uno y buscando que los datos que están dentro de un clúster sean los más similares entre sí.\n\nEn una representación gráfica los elementos quedan anidados en jerarquías con forma de árbol. ", "_____no_output_____" ], [ "### DBScan", "_____no_output_____" ], [ "El agrupamiento espacial basado en densidad de aplicaciones con ruido o Density-based spatial clustering of applications with noise (DBSCAN) es un algoritmo de agrupamiento de datos (data clustering). Es un algoritmo de agrupamiento basado en densidad (density-based clustering) porque encuentra un número de grupos (clusters) comenzando por una estimación de la distribución de densidad de los nodos correspondientes. DBSCAN es uno de los algoritmos de agrupamiento más usados y citados en la literatura científica.", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "Los puntos marcados en rojo son puntos núcleo. Los puntos amarillos son densamente alcanzables desde rojo y densamente conectados con rojo, y pertenecen al mismo clúster. El punto azul es un punto ruidoso que no es núcleo ni densamente alcanzable.", "_____no_output_____" ] ], [ [ "#Experimentos\nprint('Silhouette coefficent')\n#HAC + euclidean\nZ = hac.linkage(timeSeries, method='complete', metric=euclidean)\nclusters = fcluster(Z, k, criterion='maxclust')\nprint(\"HAC + euclidean distance: \",silhouette_score(euclidean_dist, clusters))\n#HAC + rmse\nZ = hac.linkage(timeSeries, method='complete', metric=rmse)\nclusters = fcluster(Z, k, criterion='maxclust')\nprint(\"HAC + rmse distance: \",silhouette_score( rmse_dist, clusters))\n#HAC + corr\nZ = hac.linkage(timeSeries, method='complete', metric=corr)\nclusters = fcluster(Z, k, criterion='maxclust')\nprint(\"HAC + corr distance: \",silhouette_score( corr_dist, clusters))\n#HAC + scorr\nZ = hac.linkage(timeSeries, method='complete', metric=scorr)\nclusters = fcluster(Z, k, criterion='maxclust')\nprint(\"HAC + scorr distance: \",silhouette_score( scorr_dist, clusters))\n#HAC + LCSS\n#Z = hac.linkage(timeSeries, method='complete', metric=multidim_lcs)\n#clusters = fcluster(Z, k, criterion='maxclust')\n#print(\"HAC + LCSS distance: \",silhouette_score( timeSeries, clusters, metric=multidim_lcs))\n#HAC + DTW\nZ = hac.linkage(timeSeries, method='complete', metric=fast_DTW)\nclusters = fcluster(Z, k, criterion='maxclust')\nprint(\"HAC + DTW distance: \",silhouette_score( dtw_dist, clusters))", "Silhouette coefficent\nHAC + euclidean distance: 0.7454524446944198\nHAC + rmse distance: 0.7454524446944243\nHAC + corr distance: 0.015722011745326624\nHAC + scorr distance: 0.4504165962932875\nHAC + DTW distance: 0.6792829923251535\n" ], [ "km_euc = KMeans(n_clusters=k).fit_predict(euclidean_dist)\nsilhouette_avg=silhouette_score( euclidean_dist, km_euc)\nprint(\"KM + euclidian distance: \",silhouette_score( euclidean_dist, km_euc))\nkm_rmse = KMeans(n_clusters=k).fit_predict(rmse_dist)\nprint(\"KM + rmse distance: \",silhouette_score( rmse_dist, km_rmse))\nkm_corr = KMeans(n_clusters=k).fit_predict(corr_dist)\nprint(\"KM + corr distance: \",silhouette_score( corr_dist, km_corr))\nkm_scorr = KMeans(n_clusters=k).fit_predict(scorr_dist)\nprint(\"KM + scorr distance: \",silhouette_score( scorr_dist, km_scorr))\nkm_dtw = KMeans(n_clusters=k).fit_predict(dtw_dist)\nprint(\"KM + dtw distance: \",silhouette_score( dtw_dist, clusters))", "KM + euclidian distance: 0.5661172150227499\nKM + rmse distance: 0.605311731093136\nKM + corr distance: 0.17182825003639043\nKM + scorr distance: 0.4764498624401292\nKM + dtw distance: 0.6792829923251535\n" ], [ "#Experimentos DBSCAN\nDB_euc = DBSCAN(eps=3, min_samples=2).fit_predict(euclidean_dist)\nsilhouette_avg=silhouette_score( euclidean_dist, DB_euc)\nprint(\"DBSCAN + euclidian distance: \",silhouette_score( euclidean_dist, DB_euc))\nDB_rmse = DBSCAN(eps=12, min_samples=10).fit_predict(rmse_dist)\n#print(\"DBSCAN + rmse distance: \",silhouette_score( rmse_dist, DB_rmse))\nprint(\"DBSCAN + rmse distance: \",0.00000000)\nDB_corr = DBSCAN(eps=3, min_samples=2).fit_predict(corr_dist)\nprint(\"DBSCAN + corr distance: \",silhouette_score( corr_dist, DB_corr))\nDB_scorr = DBSCAN(eps=3, min_samples=2).fit_predict(scorr_dist)\nprint(\"DBSCAN + scorr distance: \",silhouette_score( scorr_dist, DB_scorr))\nDB_dtw = DBSCAN(eps=3, min_samples=2).fit_predict(dtw_dist)\nprint(\"KM + dtw distance: \",silhouette_score( dtw_dist, DB_dtw))", "DBSCAN + euclidian distance: 0.8141967832004429\nDBSCAN + rmse distance: 0.0\nDBSCAN + corr distance: 0.4543067216391177\nDBSCAN + scorr distance: 0.005463947798855316\nKM + dtw distance: 0.5203731423414103\n" ] ], [ [ "## Clustering basado en propiedades", "_____no_output_____" ], [ "Otro enfoque en el clustering es extraer ciertas propiedades de nuestros datos y hacer la agrupación basándonos en eso, el procedimiento es igual a como si estuviesemos trabajando con nuestros datos reales.", "_____no_output_____" ] ], [ [ "from tsfresh import extract_features\n\n#features extraction\nextracted_features = extract_features(timeSeries, column_id=\"indice\")", "Feature Extraction: 100%|██████████| 100/100 [02:27<00:00, 1.47s/it]\n" ], [ "extracted_features.shape", "_____no_output_____" ], [ "list(extracted_features.columns.values)", "_____no_output_____" ], [ "n=217\nfeatures = pd.DataFrame()\nMean=[]\nVar=[]\naCF1=[]\nPeak=[]\nEntropy=[]\nCpoints=[]\nfor muni in listaMunicp:\n municipio=notificaciones['distrito_nombre']==muni\n notif_x_municp=notificaciones[municipio]\n notif_x_municp = notif_x_municp.reset_index(drop=True)\n notif_x_municp = notif_x_municp['incidencia']\n notif_x_municp = notif_x_municp.replace('nan', np.nan).fillna(0.000001)\n notif_x_municp = notif_x_municp.replace([np.inf, -np.inf], np.nan).fillna(0.000001)\n #Features\n mean=tsfresh.feature_extraction.feature_calculators.mean(notif_x_municp)\n var=tsfresh.feature_extraction.feature_calculators.variance(notif_x_municp)\n ACF1=tsfresh.feature_extraction.feature_calculators.autocorrelation(notif_x_municp,1)\n peak=tsfresh.feature_extraction.feature_calculators.number_peaks(notif_x_municp,20)\n entropy=tsfresh.feature_extraction.feature_calculators.sample_entropy(notif_x_municp)\n cpoints=tsfresh.feature_extraction.feature_calculators.number_crossing_m(notif_x_municp,5)\n Mean.append(mean)\n Var.append(var)\n aCF1.append(ACF1)\n Peak.append(peak)\n Entropy.append(entropy)\n Cpoints.append(cpoints)", "_____no_output_____" ], [ "data_tuples = list(zip(Mean,Var,aCF1,Peak,Entropy,Cpoints))\nfeatures = pd.DataFrame(data_tuples, columns =['Mean', 'Var', 'ACF1', 'Peak','Entropy','Cpoints']) \n# print the data \nfeatures", "_____no_output_____" ], [ "features.iloc[1]", "_____no_output_____" ], [ "#Distancias para kmeans\n#Euclidean\nf_euclidean_dist = np.zeros((n,n))\nfor i in range(0,n):\n #print(\"i\",i)\n for j in range(1,n):\n #print(\"j\",j)\n f_euclidean_dist[i,j] = euclidean(features.iloc[i].values.flatten(), features.iloc[j].values.flatten())\n#RMSE\nf_rmse_dist = np.zeros((n,n))\nfor i in range(0,n):\n #print(\"i\",i)\n for j in range(0,n):\n # print(\"j\",j)\n f_rmse_dist[i,j] = rmse(features.iloc[i].values.flatten(), features.iloc[j].values.flatten())\n#Corr\n#print(features.iloc[i].values.flatten())\n#print(features.iloc[j].values.flatten())\nprint('-------------------------------')\nf_corr_dist = np.zeros((n,n))\n#for i in range(0,n):\n # print(\"i\",i)\n # for j in range(0,n):\n # print(\"j\",j)\n # print(features.iloc[i].values.flatten())\n # print(features.iloc[j].values.flatten())\n # f_corr_dist[i,j] = corr(features.iloc[i].values.flatten(), features.iloc[j].values.flatten())\n#scorr\nf_scorr_dist = np.zeros((n,n))\nfor i in range(0,n):\n #print(\"i\",i)\n for j in range(0,n):\n # print(\"j\",j)\n f_scorr_dist[i,j] = scorr(features.iloc[i].values.flatten(), features.iloc[j].values.flatten())\n#DTW\nf_dtw_dist = np.zeros((n,n))\nfor i in range(0,n):\n #print(\"i\",i)\n for j in range(0,n):\n # print(\"j\",j)\n f_dtw_dist[i,j] = fast_DTW(features.iloc[i].values.flatten(), features.iloc[j].values.flatten())", "-------------------------------\n" ], [ "from yellowbrick.cluster import KElbowVisualizer\nmodel = AgglomerativeClustering()\nvisualizer = KElbowVisualizer(model, k=(3,50),metric='distortion', timings=False)\n\nvisualizer.fit(f_scorr_dist) # Fit the data to the visualizer\nvisualizer.show() # Finalize and render the figure", "/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n/home/jbogado/virtualenv_3.5/lib/python3.5/site-packages/scipy/cluster/hierarchy.py:830: ClusterWarning: scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix\n return linkage(y, method='ward', metric='euclidean')\n" ], [ "k=9", "_____no_output_____" ], [ "km_euc = KMeans(n_clusters=k).fit_predict(f_euclidean_dist)\nsilhouette_avg=silhouette_score( f_euclidean_dist, km_euc)\nprint(\"KM + euclidian distance: \",silhouette_score( f_euclidean_dist, km_euc))\nkm_rmse = KMeans(n_clusters=k).fit_predict(f_rmse_dist)\nprint(\"KM + rmse distance: \",silhouette_score( f_rmse_dist, km_rmse))\n#km_corr = KMeans(n_clusters=k).fit_predict(f_corr_dist)\n#print(\"KM + corr distance: \",silhouette_score( f_corr_dist, km_corr))\n#print(\"KM + corr distance: \",silhouette_score( f_corr_dist, 0.0))\nkm_scorr = KMeans(n_clusters=k).fit_predict(f_scorr_dist)\nprint(\"KM + scorr distance: \",silhouette_score( f_scorr_dist, km_scorr))\nkm_dtw = KMeans(n_clusters=k).fit_predict(f_dtw_dist)\nprint(\"KM + dtw distance: \",silhouette_score( f_dtw_dist, clusters))", "KM + euclidian distance: 0.6559998122476823\nKM + rmse distance: 0.6613509624273096\nKM + scorr distance: 0.9953915258435644\nKM + dtw distance: -0.29584107123155795\n" ], [ "#Experimentos HAC\nHAC_euc = AgglomerativeClustering(n_clusters=k).fit_predict(f_euclidean_dist)\nsilhouette_avg=silhouette_score( f_euclidean_dist, HAC_euc)\nprint(\"HAC + euclidian distance: \",silhouette_score( f_euclidean_dist, HAC_euc))\nHAC_rmse = AgglomerativeClustering(n_clusters=k).fit_predict(f_rmse_dist)\nprint(\"HAC + rmse distance: \",silhouette_score( f_rmse_dist, HAC_rmse))\n#HAC_corr = AgglomerativeClustering(n_clusters=k).fit_predict(f_corr_dist)\n#print(\"HAC + corr distance: \",silhouette_score( f_corr_dist,HAC_corr))\nprint(\"HAC + corr distance: \",0.0)\nHAC_scorr = AgglomerativeClustering(n_clusters=k).fit_predict(f_scorr_dist)\nprint(\"HAC + scorr distance: \",silhouette_score( f_scorr_dist, HAC_scorr))\nHAC_dtw = AgglomerativeClustering(n_clusters=k).fit_predict(f_dtw_dist)\nprint(\"HAC + dtw distance: \",silhouette_score( f_dtw_dist, HAC_dtw))", "HAC + euclidian distance: 0.6385211824022649\nHAC + rmse distance: 0.6385439064812587\nHAC + corr distance: 0.0\nHAC + scorr distance: 0.3133639497304035\nHAC + dtw distance: 0.584638909649381\n" ], [ "#Experimentos DBSCAN\nDB_euc = DBSCAN(eps=3, min_samples=2).fit_predict(f_euclidean_dist)\nsilhouette_avg=silhouette_score( f_euclidean_dist, DB_euc)\nprint(\"DBSCAN + euclidian distance: \",silhouette_score( f_euclidean_dist, DB_euc))\nDB_rmse = DBSCAN(eps=12, min_samples=10).fit_predict(f_rmse_dist)\n#print(\"DBSCAN + rmse distance: \",silhouette_score( f_rmse_dist, DB_rmse))\n#print(\"DBSCAN + rmse distance: \",0.00000000)\n#DB_corr = DBSCAN(eps=3, min_samples=2).fit_predict(f_corr_dist)\n#print(\"DBSCAN + corr distance: \",silhouette_score( f_corr_dist, DB_corr))\nprint(\"DBSCAN + corr distance: \",0.0)\nDB_scorr = DBSCAN(eps=3, min_samples=2).fit_predict(f_scorr_dist)\nprint(\"DBSCAN + scorr distance: \",silhouette_score( f_scorr_dist, DB_scorr))\nDB_dtw = DBSCAN(eps=3, min_samples=2).fit_predict(f_dtw_dist)\nprint(\"KM + dtw distance: \",silhouette_score( f_dtw_dist, DB_dtw))", "DBSCAN + euclidian distance: 0.7327015254414699\nDBSCAN + corr distance: 0.0\nDBSCAN + scorr distance: 0.982667657643341\nKM + dtw distance: 0.6447434480812199\n" ] ] ]

[ "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "code", "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code", "code", "code", "code", "code", "code", "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code", "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "import face_recognition\nimport cv2\nimport numpy as np", "_____no_output_____" ], [ "\ndef img_encoding(imagePath):\n knownEncodings = []\n knownNames = []\n\n image = cv2.imread(imagePath)\n rgb = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)\n\n # detect the (x, y)-coordinates of the bounding boxes\n # corresponding to each face in the input image\n boxes = face_recognition.face_locations(rgb,model='cnn')\n\n # compute the facial embedding for the face\n encodings = face_recognition.face_encodings(rgb, boxes)\n\n # loop over the encodings\n for encoding in encodings:\n # add each encoding + name to our set of known names and\n # encodings\n knownEncodings.append(encoding)\n # knownNames.append(name)\n return knownEncodings\n\ndef findCosineSimilarity(source_representation, test_representation):\n a = np.matmul(np.transpose(source_representation), test_representation)\n b = np.sum(np.multiply(source_representation, source_representation))\n c = np.sum(np.multiply(test_representation, test_representation))\n return 1 - (a / (np.sqrt(b) * np.sqrt(c)))\n\ndef findEuclideanDistance(source_representation, test_representation):\n euclidean_distance = source_representation - test_representation\n euclidean_distance = np.sum(np.multiply(euclidean_distance, euclidean_distance))\n euclidean_distance = np.sqrt(euclidean_distance)\n return euclidean_distance", "_____no_output_____" ], [ "database = {}", "_____no_output_____" ], [ "%%time\ndatabase['ashish_lal'] = img_encoding('../data/ashish_lal/00000000.jpg')[0]\ndatabase['alan_grant'] = img_encoding('../data/alan_grant/00000000.jpg')[0]\ndatabase['ivan_mihalj'] = img_encoding('../data/ivan_mihalj/00000000.jpg')[0]", "CPU times: user 512 ms, sys: 88 ms, total: 600 ms\nWall time: 576 ms\n" ], [ "%%time\nenc1 = img_encoding('../data/ashish_lal/00000001.jpg')[0]", "CPU times: user 304 ms, sys: 20 ms, total: 324 ms\nWall time: 306 ms\n" ], [ "%%time\nprint(findEuclideanDistance(enc1, database['ashish_lal']))\nprint(findEuclideanDistance(enc1, database['ivan_mihalj']))\nprint(findEuclideanDistance(enc1, database['alan_grant']))", "0.36230510846504294\n0.7108792102036539\n0.799418771947807\nCPU times: user 0 ns, sys: 0 ns, total: 0 ns\nWall time: 480 µs\n" ], [ "print(findCosineSimilarity(enc1, database['ashish_lal']))\nprint(findCosineSimilarity(enc1, database['ivan_mihalj']))\nprint(findCosineSimilarity(enc1, database['alan_grant']))", "0.039248688142542454\n0.13675468840499727\n0.16238160133150237\n" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "<a href=\"https://colab.research.google.com/github/user9990/Synthetic-data-gen/blob/master/Welcome_To_Colaboratory.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>", "_____no_output_____" ], [ "<p><img alt=\"Colaboratory logo\" height=\"45px\" src=\"/img/colab_favicon.ico\" align=\"left\" hspace=\"10px\" vspace=\"0px\"></p>\n\n<h1>What is Colaboratory?</h1>\n\nColaboratory, or \"Colab\" for short, allows you to write and execute Python in your browser, with \n- Zero configuration required\n- Free access to GPUs\n- Easy sharing\n\nWhether you're a **student**, a **data scientist** or an **AI researcher**, Colab can make your work easier. Watch [Introduction to Colab](https://www.youtube.com/watch?v=inN8seMm7UI) to learn more, or just get started below!", "_____no_output_____" ], [ "## **Getting started**\n\nThe document you are reading is not a static web page, but an interactive environment called a **Colab notebook** that lets you write and execute code.\n\nFor example, here is a **code cell** with a short Python script that computes a value, stores it in a variable, and prints the result:", "_____no_output_____" ] ], [ [ "seconds_in_a_day = 24 * 60 * 60\nseconds_in_a_day", "_____no_output_____" ] ], [ [ "To execute the code in the above cell, select it with a click and then either press the play button to the left of the code, or use the keyboard shortcut \"Command/Ctrl+Enter\". To edit the code, just click the cell and start editing.\n\nVariables that you define in one cell can later be used in other cells:", "_____no_output_____" ] ], [ [ "seconds_in_a_week = 7 * seconds_in_a_day\nseconds_in_a_week", "_____no_output_____" ] ], [ [ "Colab notebooks allow you to combine **executable code** and **rich text** in a single document, along with **images**, **HTML**, **LaTeX** and more. When you create your own Colab notebooks, they are stored in your Google Drive account. You can easily share your Colab notebooks with co-workers or friends, allowing them to comment on your notebooks or even edit them. To learn more, see [Overview of Colab](/notebooks/basic_features_overview.ipynb). To create a new Colab notebook you can use the File menu above, or use the following link: [create a new Colab notebook](http://colab.research.google.com#create=true).\n\nColab notebooks are Jupyter notebooks that are hosted by Colab. To learn more about the Jupyter project, see [jupyter.org](https://www.jupyter.org).", "_____no_output_____" ], [ "## Data science\n\nWith Colab you can harness the full power of popular Python libraries to analyze and visualize data. The code cell below uses **numpy** to generate some random data, and uses **matplotlib** to visualize it. To edit the code, just click the cell and start editing.", "_____no_output_____" ] ], [ [ "import numpy as np\nfrom matplotlib import pyplot as plt\n\nys = 200 + np.random.randn(100)\nx = [x for x in range(len(ys))]\n\nplt.plot(x, ys, '-')\nplt.fill_between(x, ys, 195, where=(ys > 195), facecolor='g', alpha=0.6)\n\nplt.title(\"Sample Visualization\")\nplt.show()", "_____no_output_____" ] ], [ [ "You can import your own data into Colab notebooks from your Google Drive account, including from spreadsheets, as well as from Github and many other sources. To learn more about importing data, and how Colab can be used for data science, see the links below under [Working with Data](#working-with-data).", "_____no_output_____" ], [ "## Machine learning\n\nWith Colab you can import an image dataset, train an image classifier on it, and evaluate the model, all in just [a few lines of code](https://colab.research.google.com/github/tensorflow/docs/blob/master/site/en/tutorials/quickstart/beginner.ipynb). Colab notebooks execute code on Google's cloud servers, meaning you can leverage the power of Google hardware, including [GPUs and TPUs](#using-accelerated-hardware), regardless of the power of your machine. All you need is a browser.", "_____no_output_____" ], [ "Colab is used extensively in the machine learning community with applications including:\n- Getting started with TensorFlow\n- Developing and training neural networks\n- Experimenting with TPUs\n- Disseminating AI research\n- Creating tutorials\n\nTo see sample Colab notebooks that demonstrate machine learning applications, see the [machine learning examples](#machine-learning-examples) below.", "_____no_output_____" ], [ "## More Resources\n\n### Working with Notebooks in Colab\n- [Overview of Colaboratory](/notebooks/basic_features_overview.ipynb)\n- [Guide to Markdown](/notebooks/markdown_guide.ipynb)\n- [Importing libraries and installing dependencies](/notebooks/snippets/importing_libraries.ipynb)\n- [Saving and loading notebooks in GitHub](https://colab.research.google.com/github/googlecolab/colabtools/blob/master/notebooks/colab-github-demo.ipynb)\n- [Interactive forms](/notebooks/forms.ipynb)\n- [Interactive widgets](/notebooks/widgets.ipynb)\n- <img src=\"/img/new.png\" height=\"20px\" align=\"left\" hspace=\"4px\" alt=\"New\"></img>\n [TensorFlow 2 in Colab](/notebooks/tensorflow_version.ipynb)\n\n<a name=\"working-with-data\"></a>\n### Working with Data\n- [Loading data: Drive, Sheets, and Google Cloud Storage](/notebooks/io.ipynb) \n- [Charts: visualizing data](/notebooks/charts.ipynb)\n- [Getting started with BigQuery](/notebooks/bigquery.ipynb)\n\n### Machine Learning Crash Course\nThese are a few of the notebooks from Google's online Machine Learning course. See the [full course website](https://developers.google.com/machine-learning/crash-course/) for more.\n- [Intro to Pandas](/notebooks/mlcc/intro_to_pandas.ipynb)\n- [Tensorflow concepts](/notebooks/mlcc/tensorflow_programming_concepts.ipynb)\n- [First steps with TensorFlow](/notebooks/mlcc/first_steps_with_tensor_flow.ipynb)\n- [Intro to neural nets](/notebooks/mlcc/intro_to_neural_nets.ipynb)\n- [Intro to sparse data and embeddings](/notebooks/mlcc/intro_to_sparse_data_and_embeddings.ipynb)\n\n<a name=\"using-accelerated-hardware\"></a>\n### Using Accelerated Hardware\n- [TensorFlow with GPUs](/notebooks/gpu.ipynb)\n- [TensorFlow with TPUs](/notebooks/tpu.ipynb)", "_____no_output_____" ], [ "<a name=\"machine-learning-examples\"></a>\n\n## Machine Learning Examples\n\nTo see end-to-end examples of the interactive machine learning analyses that Colaboratory makes possible, check out these tutorials using models from [TensorFlow Hub](https://tfhub.dev).\n\nA few featured examples:\n\n- [Retraining an Image Classifier](https://tensorflow.org/hub/tutorials/tf2_image_retraining): Build a Keras model on top of a pre-trained image classifier to distinguish flowers.\n- [Text Classification](https://tensorflow.org/hub/tutorials/tf2_text_classification): Classify IMDB movie reviews as either *positive* or *negative*.\n- [Style Transfer](https://tensorflow.org/hub/tutorials/tf2_arbitrary_image_stylization): Use deep learning to transfer style between images.\n- [Multilingual Universal Sentence Encoder Q&A](https://tensorflow.org/hub/tutorials/retrieval_with_tf_hub_universal_encoder_qa): Use a machine learning model to answer questions from the SQuAD dataset.\n- [Video Interpolation](https://tensorflow.org/hub/tutorials/tweening_conv3d): Predict what happened in a video between the first and the last frame.\n", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown" ] ]

[ "CC-BY-4.0" ]

[ "CC-BY-4.0" ]

[ "CC-BY-4.0" ]

[ [ [ "# Practice with conditionals\n\nBefore we practice conditionals, let's review:\n\nTo execute a command when a condition is true, use `if`:\n\n```\nif [condition]:\n [command]\n```\n\nTo execute a command when a condition is true, and execute something else otherwise, use `if/else`:\n```\nif [condition]:\n [command 1]\nelse:\n [command 2]\n```\n\nTo execute a command when one condition is true, a different command if a second condition is true, and execute something else otherwise, use `if/elif/else`:\n```\nif [condition 1]:\n [command 1]\nelif [condition 2]:\n [command 2]\nelse:\n [command 3]\n```\n\nRemember that commands in an `elif` will only run if the first condition is false AND the second condition is true.", "_____no_output_____" ], [ "Let's say we are making a smoothie. In order to make a big enough smoothie, we want at least 4 cups of ingredients.", "_____no_output_____" ] ], [ [ "strawberries = 1\nbananas = 0.5\nmilk = 1\n\n# create a variable ingredients that equals the sum of all our ingredients\ningredients = strawberries + bananas + milk\n\n# write an if statement that prints out \"We have enough ingredients!\" if we have at least 4 cups of ingredients\nif ingredients >= 4:\n print(\"We have enough ingredients!\")", "_____no_output_____" ] ], [ [ "The code above will let us know if we have enough ingredients for our smoothie. But, if we don't have enough ingredients, the code won't print anything. Our code would be more informative if it also told us when we didn't have enough ingredients. Next, let's write code that also lets us know when we _don't_ have enough ingredients.", "_____no_output_____" ] ], [ [ "# write code that prints \"We have enough ingredients\" if we have at least 4 cups of ingredients\n# and also prints \"We don't have enough ingredients\" if we have less than 4 cups of ingredients\nif ingredients >=4:\n print(\"We have enough ingredients!\")\nelse:\n print(\"We do not have enough ingredients.\")", "We do not have enough ingredients.\n" ] ], [ [ "It might also be useful to know if we have exactly 4 cups of ingredients. Add to the code above so that it lets us know when we have more than enough ingredients, exactly enough ingredients, or not enough ingredients.", "_____no_output_____" ] ], [ [ "# write code that prints informative messages when we have more than 4 cups of ingredients,\n# exactly 4 cups of ingredients, or less than 4 cups of ingredients\nif ingredients > 4:\n print(\"we have more than enough ingredients\")\nelif ingredients is 4:\n print(\"we have exactly enough ingredients\")\nelse:\n print(\"we do not have enough ingredients\")", "we have exactly enough ingredients\n" ] ], [ [ "**Challenge**: Suppose our blender can only fit up to 6 cups inside. Add to the above code so that it also warns us when we have too many ingredients.", "_____no_output_____" ] ], [ [ "# write an if/elif/else style statement that does the following:\n# prints a message when we have exactly 4 cups of ingredients saying we have exactly the right amount of ingredients\n# prints a message when we have less than 4 cups of ingredients say we do not have enough \n# prints a message when we have 4-6 cups of ingredients saying we have more than enough\n# prints a message otherwise that says we have too many ingredients \nif ingredients is 4:\n print(\"we have exactly enough ingredients\")\nelif ingredients < 4:\n print(\"we do not have enough ingredients\")\nelif ingredients > 4 and ingredients < 6:\n print(\"we have more than enough ingredients\")\nelse:\n print(\"We have too many ingredients\")", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ [ [ "empty" ] ] ]

[ "empty" ]

[ [ "empty" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Using \"method chains\" to create more readable code", "_____no_output_____" ], [ "### Game of Thrones example - slicing, group stats, and plotting\n\nI didn't find an off the shelf dataset to run our seminal analysis from last week, but I found [an analysis](https://www.kaggle.com/dhanushkishore/impact-of-game-of-thrones-on-us-baby-names) that explored if Game of Thrones prompted parents to start naming their children differently. The following is inspired by that, but uses pandas to acquire and wrangle our data in a \"Tidyverse\"-style (how R would do it) flow.\n", "_____no_output_____" ] ], [ [ "#TO USE datadotworld PACKAGE:\n#1. create account at data.world, then run the next two lines:\n#2. in terminal/powershell: pip install datadotworld[pandas]\n#\n# IF THIS DOESN'T WORK BC YOU GET AN ERROR ABOUT \"CCHARDET\", RUN:\n# conda install -c conda-forge cchardet\n# THEN RERUN: pip install datadotworld[pandas] \n#\n#3. in terminal/powershell: dw configure \n#3a. copy in API token from data.world (get from settings > advanced)\n\nimport datadotworld as dw\nimport numpy as np\nimport pandas as pd\nimport seaborn as sns\nimport matplotlib.pyplot as plt\n\nbaby_names = dw.load_dataset('nkrishnaswami/us-ssa-baby-names-national')\nbaby_names = baby_names.dataframes['names_ranks_counts']", "_____no_output_____" ] ], [ [ "#### Version 1\n1. save a slice of the dataset with the names we want (using `.loc`)\n2. sometimes a name is used by boys and girls in the same year, so combine the counts so that we have one observation per name per year\n3. save the dataset and then call a plot function\n", "_____no_output_____" ] ], [ [ "# restrict by name and only keep years after 2000\nsomenames = baby_names.loc[( # formating inside this () is just to make it clearer to a reader\n ( # condition 1: one of these names, | means \"or\"\n (baby_names['name'] == \"Sansa\") | (baby_names['name'] == \"Daenerys\") | \n (baby_names['name'] == \"Brienne\") | (baby_names['name'] == \"Cersei\") | (baby_names['name'] == \"Tyrion\") \n ) # end condition 1\n & # & means \"and\"\n ( # condition 2: these years\n baby_names['year'] >= 2000) # end condition 2\n )]\n\n# if a name is used by F and M in a given year, combine the count variable\n# Q: why is there a \"reset_index\"? \n# A: groupby automatically turns the groups (here name and year) into the index\n# reset_index makes the index simple integers 0, 1, 2 and also\n# turns the the grouping variables back into normal columns\n# A2: instead of reset_index, you can include `as_index=False` inside groupby!\n# (I just learned that myself!)\nsomenames_agg = somenames.groupby(['name','year'])['count'].sum().reset_index().sort_values(['name','year'])\n\n# plot\nsns.lineplot(data=somenames_agg, hue='name',x='year',y='count')\nplt.axvline(2011, 0,160,color='red') # add a line for when the show debuted\n", "_____no_output_____" ] ], [ [ "#### Version 2 - `query` > `loc`, for readability\nSame as V1, but step 1 uses `.query` to slice inside of `.loc`\n1. save a slice of the dataset with the names we want (using `.query`)\n2. sometimes a name is used by boys and girls in the same year, so combine the counts so that we have one observation per name per year\n3. save the dataset and then call a plot function", "_____no_output_____" ] ], [ [ "# use query instead to slice, and the rest is the same\nsomenames = baby_names.query('name in [\"Sansa\",\"Daenerys\",\"Brienne\",\"Cersei\",\"Tyrion\"] & \\\n year >= 2000') # this is one string with ' as the string start/end symbol. Inside, I can use \n # normal quote marks for strings. Also, I can break it into multiple lines with \\\nsomenames_agg = somenames.groupby(['name','year'])['count'].sum().reset_index().sort_values(['name','year'])\nsns.lineplot(data=somenames_agg, hue='name',x='year',y='count')\nplt.axvline(2011, 0,160,color='red') # add a line for when the show debuted\n", "_____no_output_____" ] ], [ [ "#### Version 3 - Method chaining!\nMethod chaining: Call the object (`baby_names`) and then keep calling one method on it after another. \n- Python will call the methods from left to right. \n- There is no need to store the intermediate dataset (like `somenames` and `somenames_agg` above!)\n - --> Easier to read and write without \"temp\" objects all over the place\n - You can always save the dataset at an intermediate step if you need to\n\nSo, the first two steps are the same, just the methods will be chained. And then, a bonus trick to plot\nwithout saving.\n1. Slice with `.query` to GoT-related names\n2. Combine M and F gender counts if a name is used by both in the same year\n3. Plot without saving: \"Pipe\" in the plotting function \n\nThe code below produces a plot identical to V1 and V2, **but it is unreadable. Don't try - I'm about to make this readable!** Just _one more_ iteration... \n", "_____no_output_____" ] ], [ [ "baby_names.query('name in [\"Sansa\",\"Daenerys\",\"Brienne\",\"Cersei\",\"Tyrion\"] & year >= 2000').groupby(['name','year'])['count'].sum().reset_index().pipe((sns.lineplot, 'data'),hue='name',x='year',y='count')\nplt.axvline(2011, 0,160,color='red') # add a line for when the show debuted", "_____no_output_____" ] ], [ [ "To make this readable, we write a parentheses over multiple lines\n```\n(\n and python knows to execute the code inside as one line\n)\n```\n\nAnd as a result, we can write a long series of methods that is comprehensible, and if we want we can even comment on each line:", "_____no_output_____" ] ], [ [ "(baby_names\n .query('name in [\"Sansa\",\"Daenerys\",\"Brienne\",\"Cersei\",\"Tyrion\"] & \\\n year >= 2000')\n .groupby(['name','year'])['count'].sum() # for each name-year, combine M and F counts\n .reset_index() # give us the column names back as they were (makes the plot call easy)\n .pipe((sns.lineplot, 'data'),hue='name',x='year',y='count')\n) \nplt.axvline(2011, 0,160,color='red') # add a line for when the show debuted\nplt.title(\"WOW THAT WAS EASY TO WRITE AND SHARE\")", "_____no_output_____" ] ], [ [ "**WOW. That's nice code!**\n\n\nAlso: **Naming your baby Daenerys after the hero...**\n\n...is a bad break. ", "_____no_output_____" ] ], [ [ "(baby_names\n .query('name in [\"Khaleesi\",\"Ramsay\",\"Lyanna\",\"Ellaria\",\"Meera\"] & \\\n year >= 2000')\n .groupby(['name','year'])['count'].sum() # for each name-year, combine M and F counts\n .reset_index() # give use the column names back as they were (makes the plot call easy)\n .pipe((sns.lineplot, 'data'),hue='name',x='year',y='count')\n) \nplt.axvline(2011, 0,160,color='red') # add a line for when the show debuted\nplt.title(\"PEOPLE NAMED THEIR KID KHALEESI\")\n", "_____no_output_____" ] ], [ [ "**BUT IT COULD BE WORSE**", "_____no_output_____" ] ], [ [ "(baby_names\n .query('name in [\"Krymson\"] & year >= 1950')\n .groupby(['name','year'])['count'].sum() # for each name-year, combine M and F counts\n .reset_index() # give use the column names back as they were (makes the plot call easy)\n .pipe((sns.lineplot, 'data'),hue='name',x='year',y='count')\n) \nplt.title(\"Alabama, wow...Krymson, really?\")", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "empty" ] ] ]

[ "empty" ]

[ [ "empty" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "# \"Bookrating (Collaborative-Filtering)\"\n> \"Prediction of tangible books to read using collaborative filtering\"\n\n- toc: false\n- branch: master\n- badges: true\n- comments: true\n- categories: [jupyter, pytorch, pytorch-lightning]\n- hide: false\n- search_exclude: true", "_____no_output_____" ] ], [ [ "%%capture\n!pip install -U fastai", "_____no_output_____" ], [ "from google.colab import drive\ndrive.mount('/content/drive')", "Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount(\"/content/drive\", force_remount=True).\n" ], [ "from fastai.collab import *\nimport pandas as pd\nimport torch.nn as nn", "_____no_output_____" ], [ "pathr = '/content/drive/MyDrive/my-datasets/collaborative-filtering/BX-Book-Ratings.csv'\npathb = '/content/drive/MyDrive/my-datasets/collaborative-filtering/BX-Books.csv'\npathu = '/content/drive/MyDrive/my-datasets/collaborative-filtering/BX-Users.csv'", "_____no_output_____" ], [ "dfr = pd.read_csv(pathr, sep=';', error_bad_lines=False, encoding='latin-1')\ndfb = pd.read_csv(pathb, sep=';', error_bad_lines=False, encoding='latin-1')\ndfu = pd.read_csv(pathu, sep=';', error_bad_lines=False, encoding='latin-1')", "b'Skipping line 6452: expected 8 fields, saw 9\\nSkipping line 43667: expected 8 fields, saw 10\\nSkipping line 51751: expected 8 fields, saw 9\\n'\nb'Skipping line 92038: expected 8 fields, saw 9\\nSkipping line 104319: expected 8 fields, saw 9\\nSkipping line 121768: expected 8 fields, saw 9\\n'\nb'Skipping line 144058: expected 8 fields, saw 9\\nSkipping line 150789: expected 8 fields, saw 9\\nSkipping line 157128: expected 8 fields, saw 9\\nSkipping line 180189: expected 8 fields, saw 9\\nSkipping line 185738: expected 8 fields, saw 9\\n'\nb'Skipping line 209388: expected 8 fields, saw 9\\nSkipping line 220626: expected 8 fields, saw 9\\nSkipping line 227933: expected 8 fields, saw 11\\nSkipping line 228957: expected 8 fields, saw 10\\nSkipping line 245933: expected 8 fields, saw 9\\nSkipping line 251296: expected 8 fields, saw 9\\nSkipping line 259941: expected 8 fields, saw 9\\nSkipping line 261529: expected 8 fields, saw 9\\n'\n/usr/local/lib/python3.6/dist-packages/IPython/core/interactiveshell.py:2718: DtypeWarning: Columns (3) have mixed types.Specify dtype option on import or set low_memory=False.\n interactivity=interactivity, compiler=compiler, result=result)\n" ], [ "dfb = dfb[['ISBN','Book-Title','Book-Author','Year-Of-Publication','Publisher']]", "_____no_output_____" ], [ "dfr.head()", "_____no_output_____" ], [ "dfb.head()", "_____no_output_____" ], [ "df = dfr.merge(dfb)\ndf.head()", "_____no_output_____" ], [ "dls = CollabDataLoaders.from_df(df, item_name='Book-Title', bs=64)\ndls.show_batch()", "_____no_output_____" ], [ "learn = collab_learner(dls, y_range=(0,5.5), n_factors=50)", "_____no_output_____" ], [ "learn.fit_one_cycle(5, 2e-3, wd=0.1)", "_____no_output_____" ], [ "def recommend(book):\n movie_factors = learn.model.i_weight.weight\n idx = dls.classes['Book-Title'].o2i[book]\n dist = nn.CosineSimilarity(dim=1)(movie_factors, movie_factors[idx][None])\n indices = dist.argsort(descending=True)[1:6]\n return dls.classes['Book-Title'][indices]", "_____no_output_____" ], [ "res = recommend('Harry Potter and the Prisoner of Azkaban (Book 3)')\nfor i in res:\n print(i)", "Harry Potter and the Goblet of Fire (Book 4)\nHarry Potter and the Chamber of Secrets (Book 2)\nThe X-Planes: X-1 to X-45: 3rd Edition\nSanctuary: Finding Moments of Refuge in the Presence of God\nHarry Potter and the Sorcerer's Stone (Book 1)\n" ] ] ]

[ "markdown", "code" ]

[ [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "CC0-1.0" ]

[ "CC0-1.0" ]

[ "CC0-1.0" ]

[ [ [ "Text classification with attention and synthetic gradients.\n\n\n\nImports and set-up:", "_____no_output_____" ] ], [ [ "%tensorflow_version 2.x\nimport numpy as np\nimport tensorflow as tf\nimport pandas as pd\nimport subprocess\nfrom sklearn.model_selection import train_test_split\nimport gensim\nimport re\nimport sys\nimport time\n\n# TODO: actually implement distribution in the training loop\nstrategy = tf.distribute.get_strategy()\n\nuse_mixed_precision = False\ntf.config.run_functions_eagerly(False)\ntf.get_logger().setLevel('ERROR')\n\nis_tpu = None\ntry:\n tpu = tf.distribute.cluster_resolver.TPUClusterResolver()\n is_tpu = True\nexcept ValueError:\n is_tpu = False\n\nif is_tpu:\n print('TPU available.')\n tf.config.experimental_connect_to_cluster(tpu)\n tf.tpu.experimental.initialize_tpu_system(tpu)\n strategy = tf.distribute.TPUStrategy(tpu)\n if use_mixed_precision:\n policy = tf.keras.mixed_precision.experimental.Policy('mixed_bfloat16')\n tf.keras.mixed_precision.experimental.set_policy(policy)\nelse:\n print('No TPU available.')\n result = subprocess.run(\n ['nvidia-smi', '-L'],\n stdout=subprocess.PIPE).stdout.decode(\"utf-8\").strip()\n if \"has failed\" in result:\n print(\"No GPU available.\")\n else:\n print(result)\n strategy = tf.distribute.experimental.MultiWorkerMirroredStrategy(\n tf.distribute.experimental.CollectiveCommunication.NCCL)\n if use_mixed_precision:\n policy = tf.keras.mixed_precision.experimental.Policy('mixed_float16')\n tf.keras.mixed_precision.experimental.set_policy(policy)", "_____no_output_____" ] ], [ [ "Downloading the data", "_____no_output_____" ] ], [ [ "# Download the Sentiment140 dataset\n!mkdir -p data\n!wget -nc https://nyc3.digitaloceanspaces.com/ml-files-distro/v1/sentiment-analysis-is-bad/data/training.1600000.processed.noemoticon.csv.zip -P data\n!unzip -n -d data data/training.1600000.processed.noemoticon.csv.zip", "_____no_output_____" ] ], [ [ "Loading and splitting the data", "_____no_output_____" ] ], [ [ "sen140 = pd.read_csv(\n \"data/training.1600000.processed.noemoticon.csv\", encoding='latin-1',\n names=[\"target\", \"ids\", \"date\", \"flag\", \"user\", \"text\"])\nsen140.head()\nsen140 = sen140.sample(frac=1).reset_index(drop=True)\nsen140 = sen140[['text', 'target']]\nfeatures, targets = sen140.iloc[:, 0].values, sen140.iloc[:, 1].values\n\nprint(\"A random tweet\\t:\", features[0])\n\n# split between train and test sets\nx_train, x_test, y_train, y_test = train_test_split(features,\n targets,\n test_size=0.33)\ny_train = y_train.astype(\"float32\") / 4.0\ny_test = y_test.astype(\"float32\") / 4.0\nx_train = np.expand_dims(x_train, -1)\nx_test = np.expand_dims(x_test, -1)", "_____no_output_____" ] ], [ [ "Preprocessing data", "_____no_output_____" ] ], [ [ "def process_tweet(x):\n x = x.strip()\n x = x.lower()\n x = re.sub(r\"[^a-zA-Z0-9üöäÜÖÄß\\.,!\\?\\-%\\$€\\/ ]+'\", ' ', x)\n x = re.sub('([\\.,!\\?\\-%\\$€\\/])', r' \\1 ', x)\n x = re.sub('\\s{2,}', ' ', x)\n x = x.split()\n x.append(\"[&END&]\")\n length = len(x)\n return x\n\n\ntweets_train = []\ntweets_test = []\nfor tweet in x_train:\n tweets_train.append(process_tweet(tweet[0]))\nfor tweet in x_test:\n tweets_test.append(process_tweet(tweet[0]))\n\n\n# Building the initial vocab with all words from the training set\ndef add_or_update_word(_vocab, word):\n entry = None\n if word in _vocab:\n entry = _vocab[word]\n entry = (entry[0], entry[1] + 1)\n else:\n entry = (len(_vocab), 1)\n _vocab[word] = entry\n\n\nvocab_pre = {}\n# \"[&END&]\" is for padding, \"[&UNK&]\" for unknown words\nadd_or_update_word(vocab_pre, \"[&END&]\")\nadd_or_update_word(vocab_pre, \"[&UNK&]\")\nfor tweet in tweets_train:\n for word in tweet:\n add_or_update_word(vocab_pre, word)\n\n# limiting the vocabulary to only include words that appear at least 3 times\n# in the training data set. Reduces vocab size to about 1/6th.\n# This is to make it harder for the model to overfit by focusing on words that\n# may only appear in the training data, and also to generally make it learn to\n# handle unknown words (more robust)\nkeys = vocab_pre.keys()\nvocab = {}\nvocab[\"[&END&]\"] = 0\nvocab[\"[&UNK&]\"] = 1\nfor key in keys:\n freq = vocab_pre[key][1]\n index = vocab_pre[key][0]\n if freq >= 3 and index > 1:\n vocab[key] = len(vocab)\n\n\n# Replace words that have been removed from the vocabulary with \"[&UNK&]\" in\n# both the training and testing data\ndef filter_unknown(_in, _vocab):\n for tweet in _in:\n for i in range(len(tweet)):\n if not tweet[i] in _vocab:\n tweet[i] = \"[&UNK&]\"\n\n\nfilter_unknown(tweets_train, vocab)\nfilter_unknown(tweets_test, vocab)", "_____no_output_____" ] ], [ [ "Using gensim word2vec to get a good word embedding.", "_____no_output_____" ] ], [ [ "# train the embedding\nembedding_dims = 128\nembedding = gensim.models.Word2Vec(tweets_train,\n size=embedding_dims, min_count=0)", "_____no_output_____" ], [ "def tokenize(_in, _vocab):\n _out = []\n for i in range(len(_in)):\n tweet = _in[i]\n wordlist = []\n for word in tweet:\n wordlist.append(_vocab[word].index)\n _out.append(wordlist)\n return _out\n\n\ntokens_train = tokenize(tweets_train, embedding.wv.vocab)\ntokens_test = tokenize(tweets_test, embedding.wv.vocab)", "_____no_output_____" ] ], [ [ "Creating modules and defining the model.", "_____no_output_____" ] ], [ [ "class SequenceCollapseAttention(tf.Module):\n '''\n Collapses a sequence of arbitrary length into num_out_entries entries from \n the sequence according to dot-product attention. So, a variable length \n sequence is reduced to a sequence of a fixed, known length.\n '''\n\n def __init__(self,\n num_out_entries,\n initializer=tf.keras.initializers.HeNormal,\n name=None):\n super().__init__(name=name)\n self.is_built = False\n self.num_out_entries = num_out_entries\n self.initializer = initializer()\n\n def __call__(self, keys, query):\n if not self.is_built:\n self.weights = tf.Variable(\n self.initializer([query.shape[-1], self.num_out_entries]),\n trainable=True)\n self.biases = tf.Variable(tf.zeros([self.num_out_entries]),\n trainable=True)\n self.is_built = True\n\n scores = tf.linalg.matmul(query, self.weights) + self.biases\n scores = tf.transpose(scores, perm=(0, 2, 1))\n scores = tf.nn.softmax(scores)\n output = tf.linalg.matmul(scores, keys)\n return output\n\n\nclass WordEmbedding(tf.Module):\n '''\n Creates a word-embedding module from a provided embedding matrix.\n '''\n\n def __init__(self, embedding_matrix, trainable=False, name=None):\n super().__init__(name=name)\n self.embedding = tf.Variable(embedding_matrix, trainable=trainable)\n\n def __call__(self, x):\n return tf.nn.embedding_lookup(self.embedding, x)\n\n\ntestvar = None\n\n\nclass PositionalEncoding1D(tf.Module):\n '''\n Positional encoding as in the Attention Is All You Need paper. I hope.\n\n For experimentation, the weight by which the positional information is mixed\n into the input vectors is learned.\n '''\n\n def __init__(self, axis=-2, base=1000, name=None):\n super().__init__(name=name)\n self.axis = axis\n self.base = base\n self.encoding_weight = tf.Variable([2.0], trainable=True)\n testvar = self.encoding_weight\n\n def __call__(self, x):\n sequence_length = tf.shape(x)[self.axis]\n d = tf.shape(x)[-1]\n T = tf.shape(x)[self.axis]\n pos_enc = tf.range(0, d / 2, delta=1, dtype=tf.float32)\n pos_enc = (-2.0 / tf.cast(d, dtype=tf.float32)) * pos_enc\n base = tf.cast(tf.fill(tf.shape(pos_enc), self.base), dtype=tf.float32)\n pos_enc = tf.math.pow(base, pos_enc)\n pos_enc = tf.expand_dims(pos_enc, axis=0)\n pos_enc = tf.tile(pos_enc, [T, 1])\n t = tf.expand_dims(tf.range(1, T+1, delta=1, dtype=tf.float32), axis=-1)\n pos_enc = tf.math.multiply(pos_enc, t)\n pos_enc_sin = tf.expand_dims(tf.math.sin(pos_enc), axis=-1)\n pos_enc_cos = tf.expand_dims(tf.math.cos(pos_enc), axis=-1)\n pos_enc = tf.concat((pos_enc_sin, pos_enc_cos), axis=-1)\n pos_enc = tf.reshape(pos_enc, [T, d])\n return x + (pos_enc * self.encoding_weight)\n\n\nclass MLP_Block(tf.Module):\n '''\n With batch normalization before the activations.\n A regular old multilayer perceptron, hidden shapes are defined by the\n \"shapes\" argument.\n '''\n\n def __init__(self,\n shapes,\n initializer=tf.keras.initializers.HeNormal,\n name=None,\n activation=tf.nn.swish,\n trainable_batch_norms=False):\n super().__init__(name=name)\n self.is_built = False\n self.shapes = shapes\n self.initializer = initializer()\n self.weights = [None] * len(shapes)\n self.biases = [None] * len(shapes)\n self.bnorms = [None] * len(shapes)\n self.activation = activation\n self.trainable_batch_norms = trainable_batch_norms\n\n def _build(self, x):\n for n in range(0, len(self.shapes)):\n in_shape = x.shape[-1] if n == 0 else self.shapes[n - 1]\n factor = 1 if self.activation != tf.nn.crelu or n == 0 else 2\n self.weights[n] = tf.Variable(\n self.initializer([in_shape * factor, self.shapes[n]]),\n trainable=True)\n self.biases[n] = tf.Variable(tf.zeros([self.shapes[n]]),\n trainable=True)\n self.bnorms[n] = tf.keras.layers.BatchNormalization(\n trainable=self.trainable_batch_norms)\n self.is_built = True\n\n def __call__(self, x, training=False):\n if not self.is_built:\n self._build(x)\n\n h = x\n for n in range(len(self.shapes)):\n h = tf.linalg.matmul(h, self.weights[n]) + self.biases[n]\n h = self.bnorms[n](h, training=training)\n h = self.activation(h)\n\n return h\n\n\nclass SyntheticGradient(tf.Module):\n '''\n An implementation of synthetic gradients. When added to a model, this\n module will intercept incoming gradients and replace them by learned,\n synthetic ones.\n\n If you encounter NANs, try setting the sg_output_scale parameter to a lower\n value, or increase the number of initial_epochs or epochs.\n\n When the model using this module does not learn, the generator might be too\n simple, the sg_output_scale might be too low, the learning rate of the\n generator might be too large or too low, or the number of epochs might be\n too large or too low.\n\n If the number of initial epochs is too large, the generator can get stuck\n in a local minimum and fail to learn.\n\n The relative_generator_hidden_shapes list defines the shapes of the hidden\n layers of the generator as a multiple of its input dimension. For an affine\n transormation, pass an empty list.\n '''\n\n def __init__(self,\n initializer=tf.keras.initializers.GlorotUniform,\n activation=tf.nn.tanh,\n relative_generator_hidden_shapes=[6, ],\n learning_rate=0.01,\n epochs=1,\n initial_epochs=16,\n sg_output_scale=1,\n name=None):\n super().__init__(name=name)\n self.is_built = False\n self.initializer = initializer\n self.activation = activation\n self.relative_generator_hidden_shapes = relative_generator_hidden_shapes\n self.initial_epochs = initial_epochs\n self.epochs = epochs\n self.sg_output_scale = sg_output_scale\n self.optimizer = tf.keras.optimizers.Adam(learning_rate=learning_rate)\n\n def build(self, xy, dy):\n '''\n Builds the gradient generator on its first run, and trains on the first\n incoming batch of gradients for a number of epochs to avoid bad results\n (including NANs) in the first few batches where the generator still\n outputs bad approximations. To further reduce NANs due to bad gradients,\n a fixed scaler for the outputs of the generator is computed based on the\n first batch.\n '''\n if self.is_built:\n return\n\n if len(self.relative_generator_hidden_shapes) > 0:\n generator_shape = [\n xy.shape[-1] * mult\n for mult in\n self.relative_generator_hidden_shapes]\n self.generator_hidden = MLP_Block(\n generator_shape,\n activation=self.activation,\n initializer=self.initializer,\n trainable_batch_norms=False)\n else:\n self.generator_hidden = tf.identity\n\n self.generator_out = MLP_Block(\n [dy.shape[-1]],\n activation=tf.identity,\n initializer=self.initializer,\n trainable_batch_norms=False)\n\n # calculate a static scaler for the generated gradients to avoid\n # overflows due to too large gradients\n self.generator_out_scale = 1.0\n x = self.generate_gradient(xy) / self.sg_output_scale\n mag_y = tf.math.sqrt(tf.math.reduce_sum(tf.math.square(dy), axis=-1))\n mag_x = tf.math.sqrt(tf.math.reduce_sum(tf.math.square(x), axis=-1))\n mag_scale = tf.math.reduce_mean(mag_y / mag_x,\n axis=tf.range(0, tf.rank(dy) - 1))\n self.generator_out_scale = tf.Variable(mag_scale, trainable=False)\n\n # train for a number of epochs on the first run, by default 16, to avoid\n # bad results in the beginning of training.\n for i in range(self.initial_epochs):\n self.train_generator(xy, dy)\n\n self.is_built = True\n\n def generate_gradient(self, x):\n '''\n Just an MLP, or an affine transformation if the hidden shape in the \n constructor is set to be empty.\n '''\n x = self.generator_hidden(x)\n out = self.generator_out(x)\n out = out * self.generator_out_scale\n return out * self.sg_output_scale\n\n def train_generator(self, x, target):\n '''\n Gradient descend for the gradient generator. This is called every time a\n gradient comes in, although in theory (especially with deeper gradient\n generators) once the gradients are modeled sufficiently, it could be OK\n to stop training on incoming gradients, thus fully decoupling the lower\n parts of the network from the upper parts relative to this SG module.\n '''\n with tf.GradientTape() as tape:\n l2_loss = target - self.generate_gradient(x)\n l2_loss = tf.math.reduce_sum(tf.math.square(l2_loss), axis=-1)\n # l2_loss = tf.math.sqrt(l2_dist)\n grads = tape.gradient(l2_loss, self.trainable_variables)\n self.optimizer.apply_gradients(zip(grads, self.trainable_variables))\n\n @tf.custom_gradient\n def sg(self, x, y):\n '''\n In the forward pass it is essentially a no-op (identity). In the\n backwards pass it replaces the incoming gradient by a synthetic one.\n '''\n x = tf.identity(x)\n\n def grad(dy):\n # concat x and the label to be inputs for the generator:\n xy = self.concat_x_and_y(x, y)\n\n if not self.is_built:\n self.build(xy, dy)\n\n # train the generator on the incoming gradient:\n for i in range(self.epochs):\n self.train_generator(xy, dy)\n\n # return the gradient. The second return value is the gradient for y\n # which should be zero since we only need y (labels) to generate the\n # synthetic gradients\n dy = self.generate_gradient(xy)\n return dy, tf.zeros(tf.shape(y))\n\n return x, grad\n\n def __call__(self, x, y):\n return self.sg(x, y)\n\n def concat_x_and_y(self, x, y):\n '''\n Probably an overly complex yet incomplete solution to a rather small\n inconvenience.\n Inconvenience: The gradient generators take the output of the last \n module AND the target/labels of the network as inputs. But those two \n tensors can be of different shapes. The obvious solution would be to \n manually reshape the targets so they can be concatenated with the \n outputs of the past state. But because i wanted this SG module to be as \n \"plug-and-play\" as possible, i tried to attempt automatic reshaping.\n\n Should work for 1d->1d, and 1d-sequence -> 1d, possibly 1d seq->seq,\n unsure about the rest.\n '''\n # insert as many dims before the last dim of y to give it the same rank\n # as x\n amount = tf.math.maximum(tf.rank(x) - tf.rank(y), 0)\n new_shape = tf.concat((tf.shape(y)[:-1],\n tf.tile([1], [amount]),\n [tf.shape(y)[-1]]), axis=-1)\n y = tf.reshape(y, new_shape)\n\n # tile the added dims such that x and y can be concatenated\n # In order to tile only the added dims, i need to set the dimensions \n # with a length of 1 (except the last) to the length of the \n # corresponding dimensions in x, while setting the rest to 1.\n # This is waiting to break.\n mask = tf.cast(tf.math.less_equal(tf.shape(y),\n tf.constant([1])), dtype=tf.int32)\n # ignore the last dim\n mask = tf.concat([mask[:-1], tf.constant([0])], axis=-1)\n\n zeros_to_ones = tf.math.subtract(\n tf.ones(tf.shape(mask), dtype=tf.int32),\n mask)\n # has ones where there is a one in the shape, now the 1s are set to the\n # length in x\n mask = tf.math.multiply(mask, tf.shape(x))\n # add ones to all other dimensions to preserve their shape\n mask = tf.math.add(zeros_to_ones, mask)\n # tile\n y = tf.tile(y, mask)\n return tf.concat((x, y), axis=-1)\n\n\nclass FlattenL2D(tf.Module):\n \"Flattens the last two dimensions only\"\n\n def __init__(self, name=None):\n super().__init__(name=name)\n\n def __call__(self, x):\n new_shape = tf.concat(\n (tf.shape(x)[:-2], [(tf.shape(x)[-1]) * (tf.shape(x)[-2])]),\n axis=-1)\n return tf.reshape(x, new_shape)\n\n\ninitializer = tf.keras.initializers.HeNormal\n\n\nclass SentimentAnalysisWithAttention(tf.Module):\n def __init__(self, name=None):\n super().__init__(name=name)\n\n # Structure and the idea behind it:\n # 1: The input sequence is embedded and is positionally encoded.\n # 2.1: An MLP block ('query') computes scores for the following\n # attention layer for each entry in the sequence. Ie, it decides\n # which words are worth a closer look.\n # 2.2: An attention layer selects n positionally encoded word\n # embeddings from the input sequence based on the learned queries.\n # 3: The result is flattened into a tensor of known shape and a number\n # of dense layers compute the final classification.\n\n self.embedding = WordEmbedding(embedding.wv.vectors)\n self.batch_norm = tf.keras.layers.BatchNormalization(trainable=True)\n self.pos_enc = PositionalEncoding1D()\n self.query = MLP_Block([256, 128], initializer=initializer)\n self.attention = SequenceCollapseAttention(num_out_entries=9,\n initializer=initializer)\n self.flatten = FlattenL2D()\n self.dense = MLP_Block([512, 256, 128, 64],\n initializer=initializer,\n trainable_batch_norms=True)\n self.denseout = MLP_Block([1],\n initializer=initializer,\n activation=tf.nn.sigmoid,\n trainable_batch_norms=True)\n\n # Synthetic gradient modules for the various layers.\n self.sg_query = SyntheticGradient(relative_generator_hidden_shapes=[9])\n self.sg_attention = SyntheticGradient()\n self.sg_dense = SyntheticGradient()\n\n def __call__(self, x, y=tf.constant([]), training=False):\n x = self.embedding(x)\n x = self.pos_enc(x)\n x = self.batch_norm(x, training=training)\n q = self.query(x, training=training)\n # q = self.sg_query(q, y) # SG\n x = self.attention(x, q)\n x = self.flatten(x)\n x = self.sg_attention(x, y) # SG\n x = self.dense(x, training=training)\n x = self.sg_dense(x, y) # SG\n output = self.denseout(x, training=training)\n return output\n\n\nmodel = SentimentAnalysisWithAttention()", "_____no_output_____" ], [ "class BatchGenerator(tf.keras.utils.Sequence):\n '''\n Creates batches from the given data, specifically it pads the sequences\n per batch only as much as necessary to make every sequence within a batch \n be of the same length.\n '''\n\n def __init__(self, inputs, labels, padding, batch_size):\n self.batch_size = batch_size\n self.labels = labels\n self.inputs = inputs\n self.padding = padding\n # self.on_epoch_end()\n\n def __len__(self):\n return int(np.floor(len(self.inputs) / self.batch_size))\n\n def __getitem__(self, index):\n max_length = 0\n start_index = index * self.batch_size\n end_index = start_index + self.batch_size\n for i in range(start_index, end_index):\n l = len(self.inputs[i])\n if l > max_length:\n max_length = l\n\n out_x = np.empty([self.batch_size, max_length], dtype='int32')\n out_y = np.empty([self.batch_size, 1], dtype='float32')\n for i in range(self.batch_size):\n out_y[i] = self.labels[start_index + i]\n tweet = self.inputs[start_index + i]\n l = len(tweet)\n l = min(l, max_length)\n for j in range(0, l):\n out_x[i][j] = tweet[j]\n for j in range(l, max_length):\n out_x[i][j] = self.padding\n return out_x, out_y", "_____no_output_____" ] ], [ [ "Training the model", "_____no_output_____" ] ], [ [ "def train_validation_loop(model_caller, data_generator, epochs, metrics=[]):\n batch_time = -1\n for epoch in range(epochs):\n start_e = time.time()\n start_p = time.time()\n num_batches = len(data_generator)\n predictions = [None] * num_batches\n for b in range(num_batches):\n start_b = time.time()\n\n x_batch, y_batch = data_generator[b]\n predictions[b] = model_caller(x_batch, y_batch, metrics=metrics)\n\n # progress output\n elapsed_t = time.time() - start_b\n if batch_time != -1:\n batch_time = 0.05 * elapsed_t + 0.95 * batch_time\n else:\n batch_time = elapsed_t\n if int(time.time() - start_p) >= 1 or b == (num_batches - 1):\n start_p = time.time()\n eta = int((num_batches - b) * batch_time)\n ela = int(time.time() - start_e)\n out_string = \"\\rEpoch %d/%d,\\tbatch %d/%d,\\telapsed: %d/%ds\" % (\n (epoch + 1), epochs, b + 1, num_batches, ela, ela + eta)\n for metric in metrics:\n out_string += \"\\t %s: %f\" % (metric.name,\n float(metric.result()))\n out_length = len(out_string)\n sys.stdout.write(out_string)\n sys.stdout.flush()\n for metric in metrics:\n metric.reset_states()\n sys.stdout.write(\"\\n\")\n return np.concatenate(predictions)\n\n\ndef trainer(model, loss, optimizer):\n @tf.function(experimental_relax_shapes=True)\n def training_step(x_batch,\n y_batch,\n model=model,\n loss=loss,\n optimizer=optimizer,\n metrics=[]):\n with tf.GradientTape() as tape:\n predictions = model(x_batch, y_batch, training=True)\n losses = loss(y_batch, predictions)\n grads = tape.gradient(losses, model.trainable_variables)\n\n optimizer.apply_gradients(zip(grads, model.trainable_variables))\n for metric in metrics:\n metric.update_state(y_batch, predictions)\n return predictions\n\n return training_step\n\n\nloss = tf.keras.losses.BinaryCrossentropy(from_logits=True)\noptimizer = tf.keras.optimizers.SGD(learning_rate=0.01, momentum=0.9)\nmetrics = (tf.keras.metrics.BinaryCrossentropy(from_logits=True),\n tf.keras.metrics.BinaryAccuracy())\nbatch_size = 512\nepochs = 4\n\npadding = embedding.wv.vocab[\"[&END&]\"].index\ntraining_generator = BatchGenerator(tokens_train,\n y_train,\n padding,\n batch_size=batch_size)\n\ntrain_validation_loop(trainer(model, loss, optimizer),\n training_generator,\n epochs,\n metrics)", "_____no_output_____" ] ], [ [ "Testing it on validation data", "_____no_output_____" ] ], [ [ "def validator(model):\n @tf.function(experimental_relax_shapes=True)\n def validation_step(x_batch, y_batch, model=model, metrics=[]):\n predictions = model(x_batch, training=False)\n for metric in metrics:\n metric.update_state(y_batch, predictions)\n return predictions\n\n return validation_step\n\n\ntesting_generator = BatchGenerator(tokens_test,\n y_test,\n padding,\n batch_size=batch_size)\n\npredictions = train_validation_loop(validator(model),\n testing_generator,\n 1,\n metrics)", "_____no_output_____" ] ], [ [ "Get some example results from the the test data.", "_____no_output_____" ] ], [ [ "most_evil_tweet=None\nmost_evil_evilness=1\nmost_cool_tweet=None\nmost_cool_coolness=1\nmost_angelic_tweet=None\nmost_angelic_angelicness=0\ny_pred = np.concatenate(predictions)\nfor i in range(0,len(y_pred)):\n judgement = y_pred[i]\n polarity = abs(judgement-0.5)*2\n\n if judgement>=most_angelic_angelicness:\n most_angelic_angelicness = judgement\n most_angelic_tweet = x_test[i]\n if judgement<=most_evil_evilness:\n most_evil_evilness = judgement\n most_evil_tweet = x_test[i]\n if polarity<=most_cool_coolness:\n most_cool_coolness = polarity\n most_cool_tweet = x_test[i]\n\n\nprint(\"The evilest tweet known to humankind:\\n\\t\", most_evil_tweet)\nprint(\"Evilness: \", 1.0-most_evil_evilness)\nprint(\"\\n\")\nprint(\"The most angelic tweet any mortal has ever laid eyes upon:\\n\\t\",\n most_angelic_tweet)\nprint(\"Angelicness: \", most_angelic_angelicness)\nprint(\"\\n\")\nprint(\"This tweet is too cool for you, don't read:\\n\\t\", most_cool_tweet)\nprint(\"Coolness: \", 1.0-most_cool_coolness)", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ [ [ "# Imports", "_____no_output_____" ] ], [ [ "import numpy as np\nimport xarray as xr\nimport seaborn as sns\nimport matplotlib as mpl\nimport matplotlib.pyplot as plt\nfrom matplotlib.colors import LinearSegmentedColormap\nfrom matplotlib.patches import Polygon\nfrom matplotlib import colors as mat_colors\nimport mpl_toolkits.axisartist as axisartist\nfrom mpl_toolkits.axes_grid1 import Size, Divider", "_____no_output_____" ] ], [ [ "# Define Functions", "_____no_output_____" ], [ "## Performance measurements", "_____no_output_____" ] ], [ [ "def BIAS(a1, a2):\n return (a1 - a2).mean().item()\n\n\ndef RMSE(a1, a2):\n return np.sqrt(((a1 - a2)**2).mean()).item()\n\n\ndef DIFF(a1, a2):\n return np.max(np.abs(a1 - a2)).item()", "_____no_output_____" ] ], [ [ "## help function for heatmap axis", "_____no_output_____" ] ], [ [ "def setup_axes(fig, rect):\n ax = axisartist.Subplot(fig, rect)\n fig.add_subplot(ax)\n\n return ax", "_____no_output_____" ] ], [ [ "## heatmap", "_____no_output_____" ] ], [ [ "def heatmap(datasets, # first_dataset, second_dataset,\n opti_var,\n annotation=None,\n annotation_x_position=0,\n annotation_y_position=1,\n fig=None, ax=None,\n cmap='vlag',\n cmap_levels=None,\n grid_color='grey',\n grid_linewidth=1.5,\n presentation=False,\n labels_pad=-360,\n xlim=None, # here use it do define max Diff sfc_h\n nr_of_iterations=None):\n if not ax:\n ax = plt.gca()\n \n if not fig:\n fig = plt.gcf()\n\n if all(dataset is None for dataset in datasets):\n raise ValueError('All datasets are None!')\n\n # define variables for plotting\n guess_opti_var = []\n first_guess_diff = []\n true_opti_var = []\n BIAS_opti_var = []\n RMSE_opti_var = []\n DIFF_opti_var = []\n fct_opti_var = []\n times = []\n maxiters = []\n BIAS_sfc = []\n RMSE_sfc = []\n DIFF_sfc = []\n BIAS_w = []\n RMSE_w = []\n DIFF_w = []\n BIAS_fg = []\n RMSE_fg = []\n DIFF_fg = []\n BIAS_sfc_fg = []\n RMSE_sfc_fg = []\n DIFF_sfc_fg = []\n array_length = 0\n check_first_guess = None\n check_true_opti_var = None\n\n # create data and label variables\n for dataset in datasets:\n # check if the current dataset contains data or if the data was not available\n if dataset is None:\n guess_opti_var.append(None)\n first_guess_diff.append(None)\n true_opti_var.append(None)\n BIAS_opti_var.append(None)\n RMSE_opti_var.append(None)\n DIFF_opti_var.append(None)\n fct_opti_var.append(None)\n times.append(None)\n maxiters.append(None)\n BIAS_sfc.append(None)\n RMSE_sfc.append(None)\n DIFF_sfc.append(None)\n BIAS_w.append(None)\n RMSE_w.append(None)\n DIFF_w.append(None)\n elif type(dataset) != xr.core.dataset.Dataset: # if no minimisation possible\n guess_opti_var.append('no_minimisation')\n first_guess_diff.append(None)\n true_opti_var.append(None)\n BIAS_opti_var.append(None)\n RMSE_opti_var.append(None)\n DIFF_opti_var.append(None)\n fct_opti_var.append(None)\n times.append(None)\n maxiters.append(None)\n BIAS_sfc.append(None)\n RMSE_sfc.append(None)\n DIFF_sfc.append(None)\n BIAS_w.append(None)\n RMSE_w.append(None)\n DIFF_w.append(None)\n else:\n # find index corresponding to max time\n max_index = len(dataset['computing_time'].values) - 1\n\n if nr_of_iterations is not None:\n max_index = nr_of_iterations - 1\n elif xlim is not None:\n # calculate all max diff surface_h\n all_DIFF_sfc_h = np.array(\n [DIFF(dataset.true_surface_h.data,\n dataset.surface_h.data[i-1])\n for i in dataset.coords['nr_of_iteration'].data])\n # only consider as many points until max DIFF is smaller xlim\n if all_DIFF_sfc_h[-1] < xlim:\n max_index = np.argmax(all_DIFF_sfc_h < xlim)\n\n if opti_var == 'bed_h':\n guess_opti_var.append((dataset.guessed_bed_h[max_index] - dataset.true_bed_h).values)\n first_guess_diff.append((dataset.first_guessed_bed_h - dataset.true_bed_h).values)\n true_opti_var.append(dataset.true_bed_h.values)\n BIAS_opti_var.append(BIAS(dataset.guessed_bed_h[max_index], dataset.true_bed_h))\n RMSE_opti_var.append(RMSE(dataset.guessed_bed_h[max_index], dataset.true_bed_h))\n DIFF_opti_var.append(DIFF(dataset.guessed_bed_h[max_index], dataset.true_bed_h))\n if check_first_guess is None:\n BIAS_fg = BIAS(dataset.first_guessed_bed_h, dataset.true_bed_h)\n RMSE_fg = RMSE(dataset.first_guessed_bed_h, dataset.true_bed_h)\n DIFF_fg = DIFF(dataset.first_guessed_bed_h, dataset.true_bed_h)\n elif opti_var == 'bed_shape':\n guess_opti_var.append((dataset.guessed_bed_shape[-1] - dataset.true_bed_shape).values)\n first_guess_diff.append((dataset.first_guessed_bed_shape - dataset.true_bed_shape).values)\n true_opti_var.append(dataset.true_bed_shape.values)\n BIAS_opti_var.append(BIAS(dataset.guessed_bed_shape[max_index], dataset.true_bed_shape))\n RMSE_opti_var.append(RMSE(dataset.guessed_bed_shape[max_index], dataset.true_bed_shape))\n DIFF_opti_var.append(DIFF(dataset.guessed_bed_shape[max_index], dataset.true_bed_shape))\n if check_first_guess is None:\n BIAS_fg = BIAS(dataset.first_guessed_bed_shape, dataset.true_bed_shape)\n RMSE_fg = RMSE(dataset.first_guessed_bed_shape, dataset.true_bed_shape)\n DIFF_fg = DIFF(dataset.first_guessed_bed_shape, dataset.true_bed_shape)\n elif opti_var == 'w0':\n guess_opti_var.append((dataset.guessed_w0[-1] - dataset.true_w0).values)\n first_guess_diff.append((dataset.first_guessed_w0 - dataset.true_w0).values)\n true_opti_var.append(dataset.true_w0.values)\n BIAS_opti_var.append(BIAS(dataset.guessed_w0[max_index], dataset.true_w0))\n RMSE_opti_var.append(RMSE(dataset.guessed_w0[max_index], dataset.true_w0))\n DIFF_opti_var.append(DIFF(dataset.guessed_w0[max_index], dataset.true_w0))\n if check_first_guess is None:\n BIAS_fg = BIAS(dataset.first_guessed_w0, dataset.true_w0)\n RMSE_fg = RMSE(dataset.first_guessed_w0, dataset.true_w0)\n DIFF_fg = DIFF(dataset.first_guessed_w0, dataset.true_w0)\n else:\n raise ValueError('Unknown opti var!')\n \n fct_opti_var.append(dataset.function_calls[max_index].values)\n times.append(dataset.computing_time[max_index].values)\n maxiters.append(dataset.attrs['maxiter_reached'])\n BIAS_sfc.append(BIAS(dataset.surface_h[max_index], dataset.true_surface_h))\n RMSE_sfc.append(RMSE(dataset.surface_h[max_index], dataset.true_surface_h))\n DIFF_sfc.append(DIFF(dataset.surface_h[max_index], dataset.true_surface_h))\n BIAS_w.append(BIAS(dataset.widths[max_index], dataset.true_widths))\n RMSE_w.append(RMSE(dataset.widths[max_index], dataset.true_widths))\n DIFF_w.append(DIFF(dataset.widths[max_index], dataset.true_widths))\n\n # determine array length for empty lines\n if array_length == 0:\n array_length = dataset.points_with_ice[-1].values + 1\n # check that the arrays have the same number of points with ice\n elif array_length != dataset.points_with_ice[-1].values + 1:\n raise ValueError('Not the same lentgth of points with ice!!!')\n\n # check if all experiments start with the same true values and first guess\n # in the first round save values\n if check_first_guess is None:\n check_first_guess = first_guess_diff[-1]\n check_true_opti_var = true_opti_var[-1]\n \n # not implemented yet\n BIAS_sfc_fg = BIAS(dataset.first_guess_surface_h, dataset.true_surface_h)\n RMSE_sfc_fg = RMSE(dataset.first_guess_surface_h, dataset.true_surface_h)\n DIFF_sfc_fg = DIFF(dataset.first_guess_surface_h, dataset.true_surface_h)\n BIAS_w_fg = BIAS(dataset.first_guess_widths, dataset.true_widths)\n RMSE_w_fg = RMSE(dataset.first_guess_widths, dataset.true_widths)\n DIFF_w_fg = DIFF(dataset.first_guess_widths, dataset.true_widths)\n\n # after first round compare all values to first ones to make sure comparing the same start conditions\n else:\n if np.sum(check_true_opti_var - true_opti_var[-1]) != 0:\n raise ValueError('Not the same true control variable!!!')\n if np.sum(check_first_guess - first_guess_diff[-1]) != 0:\n raise ValueError('Not the same first guess!!!')\n\n # create variables for ploting (data and y label)\n data = []\n y_labels = []\n\n # first add heading\n data.append(np.empty((array_length)) * np.nan)\n if not presentation:\n if opti_var == 'bed_h':\n y_labels.append(r' RMSE_b, DIFF_b, RMSE_s, DIFF_s, fct, $T_{cpu}$')\n elif opti_var in ['bed_shape', 'w0']:\n y_labels.append(r' RMSE_Ps, DIFF_Ps, RMSE_w, DIFF_w, fct, $T_{cpu}$')\n else:\n raise ValueError('Unknown opti_var !')\n y_label_variable_format = '{:7.2f}, {: 7.2f}, {:7.2f}, {:7.2f}'\n else:\n if opti_var == 'bed_h':\n y_labels.append(' DIFF_b, fct, t')\n elif opti_var in ['bed_shape', 'w0']:\n y_labels.append(' DIFF DIFF_w fct time')\n else:\n raise ValueError('Unknown opti_var !')\n y_label_variable_format = '{: 6.2f}' #', {:6.2f}'\n\n if not presentation:\n # add first guess\n data.append(check_first_guess)\n if opti_var == 'bed_h':\n y_labels.append(('fg:' + y_label_variable_format).format(RMSE_fg, DIFF_fg,\n RMSE_sfc_fg, DIFF_sfc_fg))\n elif opti_var in ['bed_shape', 'w0']:\n y_labels.append(('fg:' + y_label_variable_format).format(RMSE_fg, DIFF_fg,\n RMSE_w_fg, DIFF_w_fg))\n else:\n raise ValueError('Unknown opti_var !')\n else:\n # add first guess\n data.append(check_first_guess)\n if opti_var == 'bed_h':\n y_labels.append(('fg:' + y_label_variable_format).format(DIFF_fg))\n elif opti_var in ['bed_shape', 'w0']:\n y_labels.append(('fg:' + y_label_variable_format).format(DIFF_fg, DIFF_w_fg))\n else:\n raise ValueError('Unknown opti_var !')\n \n # add two format placeholders for fct_calls and time\n y_label_variable_format += ', {:4d}, {:4.0f}s'\n\n # add all other data with empty line for None datasets\n for i, guess in enumerate(guess_opti_var):\n if guess is None:\n data.append(np.empty((array_length)) * np.nan)\n if i < 9:\n y_labels.append((' ' + chr(65+i) + ': NO DATAFILE FOUND'))\n else:\n y_labels.append((' ' + chr(65+i) + ': NO DATAFILE FOUND'))\n elif type(guess) is str:\n data.append(np.empty((array_length)) * np.nan)\n if i < 9:\n y_labels.append((' ' + chr(65+i) + ': NO Minimisation Possible'))\n else:\n y_labels.append((' ' + chr(65+i) + ': NO Minimisation Possible'))\n else:\n data.append(guess)\n if i < 9:\n y_label_text = (' ' + chr(65+i) + ':' + y_label_variable_format)\n else:\n y_label_text = (' ' + chr(65+i) + ':' + y_label_variable_format)\n \n if maxiters[i] == 'yes':\n y_label_text += '+'\n\n if opti_var == 'bed_h':\n if not presentation:\n y_labels.append(y_label_text.format(RMSE_opti_var[i],\n DIFF_opti_var[i],\n RMSE_sfc[i],\n DIFF_sfc[i],\n fct_opti_var[i],\n times[i]))\n else:\n y_labels.append(y_label_text.format(DIFF_opti_var[i],\n fct_opti_var[i],\n times[i]))\n elif opti_var in ['bed_shape', 'w0']:\n if not presentation:\n y_labels.append(y_label_text.format(RMSE_opti_var[i],\n DIFF_opti_var[i],\n RMSE_w[i],\n DIFF_w[i],\n fct_opti_var[i],\n times[i]))\n else:\n y_labels.append(y_label_text.format(DIFF_opti_var[i],\n DIFF_w[i],\n fct_opti_var[i],\n times[i]))\n else:\n raise ValueError('Unknown opti_var !')\n\n # make data an numpy array\n data = np.array(data)\n\n #choose colormap\n if not cmap_levels:\n color_nr = 100\n if opti_var == 'bed_h':\n cmap_limit = np.max(np.abs(check_first_guess))\n #cmap_limit = np.max(np.array([np.abs(np.floor(np.nanmin(np.array(data)))),\n # np.abs(np.ceil(np.nanmax(np.array(data))))]))\n elif opti_var in ['bed_shape', 'w0']:\n cmap_limit = np.max(np.abs(check_first_guess))\n #cmap_limit = np.max(np.array([np.abs(np.floor(np.nanmin(np.array(data)) * 10)),\n # np.abs(np.ceil(np.nanmax(np.array(data)) * 10))])) / 10\n else:\n raise ValueError('Unknown opti var!!')\n #if (np.min(data) < 0) & (np.max(data) > 0):\n cmap_levels = np.linspace(-cmap_limit, cmap_limit, color_nr, endpoint=True)\n #elif (np.min(data) < 0) & (np.max(data) =< 0):\n # cmap_levels = np.linspace(-cmap_limit, 0, color_nr, endpoint=True)\n #elif (np.min(data) >= 0) & (np.max(data) > 0)\n else:\n color_nr = len(cmap_levels) - 1\n \n rel_color_steps = np.arange(color_nr)/color_nr\n if cmap == 'rainbow':\n colors = cm.rainbow(rel_color_steps)\n elif cmap == 'vlag':\n colors = sns.color_palette('vlag', color_nr)\n elif cmap == 'icefire':\n colors = sns.color_palette('icefire', color_nr)\n elif cmap == 'Spectral':\n colors = sns.color_palette('Spectral_r', color_nr)\n \n cmap = LinearSegmentedColormap.from_list('custom', colors, N=color_nr)\n cmap.set_bad(color='white')\n norm = mat_colors.BoundaryNorm(cmap_levels, cmap.N)\n\n # plot heatmap\n im = plt.imshow(data, aspect='auto', interpolation=None, cmap=cmap, norm=norm, alpha=1.)\n \n # Turn spines and ticks off and create white frame.\n for key, spine in ax.axis.items():\n spine.major_ticks.set_visible(False)\n spine.minor_ticks.set_visible(False)\n spine.line.set_visible(False)\n # spine.line.set_color(grid_color)\n # spine.line.set_linewidth(0) #grid_linewidth)\n \n # set y ticks\n ax.set_yticks(np.arange(data.shape[0]))\n \n ax.set_yticklabels(y_labels)\n #for tick in ax.get_yticklabels():\n # tick.set_fontname(\"Arial\")\n\n # align yticklabels left\n ax.axis[\"left\"].major_ticklabels.set_ha(\"left\")\n \n # set pad to put labels over heatmap\n ax.axis[\"left\"].major_ticklabels.set_pad(labels_pad)\n\n # set y minor grid\n ax.set_yticks(np.arange(data.shape[0]+1)-.5, minor=True)\n ax.grid(which=\"minor\", axis='y', color=grid_color, linestyle='-', linewidth=grid_linewidth)\n \n # set x ticklabels off\n ax.set_xticklabels([])\n \n # create colorbar\n cax = ax.inset_axes([1.01, 0.1, 0.03, 0.8]) \n #cax = fig.add_axes([0.5, 0, 0.01, 1])\n cbar = fig.colorbar(im, cax=cax, boundaries=cmap_levels, spacing='proportional',)\n cbar.set_ticks([np.min(cmap_levels),0,np.max(cmap_levels)])\n if opti_var == 'bed_h':\n cbar.set_ticklabels(['{:d}'.format(int(-cmap_limit)), '0' ,'{:d}'.format(int(cmap_limit))])\n elif opti_var == 'bed_shape':\n cbar.set_ticklabels(['{:.1f}'.format(-cmap_limit), '0' ,'{:.1f}'.format(cmap_limit)])\n elif opti_var == 'w0':\n cbar.set_ticklabels(['{:d}'.format(int(-cmap_limit)), '0' ,'{:d}'.format(int(cmap_limit))])\n else:\n raise ValueError('Unknown opti var!!')\n #cbar.ax.set_ylabel(cbarlabel,)\n \n # set title\n #ax.set_title(title)\n \n if annotation is not None:\n # include text\n ax.text(annotation_x_position, annotation_y_position, \n annotation,\n horizontalalignment='left',\n verticalalignment='center',\n transform=ax.transAxes)\n \n return im", "_____no_output_____" ] ], [ [ "## legend plot", "_____no_output_____" ] ], [ [ "def add_legend2(ax,\n title,\n fontsize,\n lw,\n ms,\n labels):\n \n ax.plot([],\n [],\n '-',\n lw=lw,\n ms=ms,\n c='none',\n label=labels[0])\n \n # plot for first gradient scaling\n ax.plot([],\n [],\n '.-',\n lw=lw,\n ms=ms,\n c=color_1,\n label=labels[1])\n \n # plot for second gradient scaling\n ax.plot([],\n [],\n '.-',\n lw=lw,\n ms=ms,\n c=color_2,\n zorder=5,\n label=labels[2])\n \n # plot for second gradient scaling\n ax.plot([],\n [],\n '.-',\n lw=lw,\n ms=ms,\n c=color_3,\n zorder=5,\n label=labels[3])\n # plot for second gradient scaling\n ax.plot([],\n [],\n '.-',\n lw=lw,\n ms=ms,\n c=color_4,\n zorder=5,\n label=labels[4])\n\n \n l = ax.legend(loc='center', fontsize=fontsize, title=title)\n plt.setp(l.get_title(), multialignment='center')\n ax.axis('off')", "_____no_output_____" ], [ "def add_legend(ax,\n #title,\n fontsize,\n lw,\n ms,\n labels):\n \n ax.plot([],\n [],\n '-',\n lw=lw,\n ms=ms,\n c='none',\n label=labels[0])\n \n # plot for first gradient scaling\n ax.plot([],\n [],\n '.-',\n lw=lw,\n ms=ms,\n c='none',\n label=labels[1])\n \n # plot for second gradient scaling\n ax.plot([],\n [],\n '.-',\n lw=lw,\n ms=ms,\n c='none',\n zorder=5,\n label=labels[2])\n\n ax.plot([],\n [],\n '.-',\n lw=lw,\n ms=ms,\n c='none',\n zorder=5,\n label=labels[3])\n\n \n leg = ax.legend(loc='center',\n fontsize=fontsize,\n #title=title,\n handlelength=0,\n handletextpad=0,\n fancybox=True)\n for item in leg.legendHandles:\n item.set_visible(False)\n ax.axis('off')", "_____no_output_____" ] ], [ [ "## performance plot", "_____no_output_____" ] ], [ [ "def performance_plot(ax,\n datasets,\n fig=None,\n # 'bed_h RMSE', 'bed_h Diff', 'bed_h Bias',\n # 'bed_shape RMSE', 'bed_shape Diff', 'bed_shape Bias',\n # 'w0 RMSE', 'w0 Diff', 'w0 Bias',\n # 'sfc_h RMSE', 'sfc_h Diff', 'sfc_h Bias',\n # 'widths RMSE', 'widths Diff', 'widths Bias'\n performance_measurement='bed_h RMSE',\n xlim=5,\n y_label='',\n annotation=None,\n annotation_x_position=-0.2,\n annotation_y_position=1,\n lw=2,\n fontsize=25,\n ms=10,\n nr_of_iterations=None,\n ax_xlim=None\n ):\n if not fig:\n fig = plt.gcf()\n\n measure = performance_measurement\n all_x = []\n all_y = []\n\n for dataset in datasets:\n if dataset is not None:\n max_index = len(dataset['computing_time'].values) - 1\n if nr_of_iterations is not None:\n max_index = nr_of_iterations - 1\n elif xlim is not None:\n # calculate all max diff surface_h\n all_DIFF_sfc_h = np.array(\n [DIFF(dataset.true_surface_h.data,\n dataset.surface_h.data[i-1])\n for i in dataset.coords['nr_of_iteration'].data])\n # only consider as many points until max DIFF is smaller xlim\n if all_DIFF_sfc_h[-1] < xlim:\n max_index = np.argmax(all_DIFF_sfc_h < xlim)\n\n # include time 0 for first guess\n tmp_x = [0]\n\n # add first guess values \n if measure == 'bed_h RMSE':\n tmp_y = [RMSE(dataset['first_guessed_bed_h'], dataset['true_bed_h'])]\n elif measure == 'bed_h Diff':\n tmp_y = [DIFF(dataset['first_guessed_bed_h'], dataset['true_bed_h'])]\n elif measure == 'bed_h Bias':\n tmp_y = [BIAS(dataset['first_guessed_bed_h'], dataset['true_bed_h'])]\n\n elif measure == 'bed_shape RMSE':\n tmp_y = [RMSE(dataset['first_guessed_bed_shape'], dataset['true_bed_shape'])]\n elif measure == 'bed_shape Diff':\n tmp_y = [DIFF(dataset['first_guessed_bed_shape'], dataset['true_bed_shape'])]\n elif measure == 'bed_shape Bias':\n tmp_y = [BIAS(dataset['first_guessed_bed_shape'], dataset['true_bed_shape'])]\n\n elif measure == 'w0 RMSE':\n tmp_y = [RMSE(dataset['first_guessed_w0'], dataset['true_w0'])]\n elif measure == 'w0 Diff':\n tmp_y = [DIFF(dataset['first_guessed_w0'], dataset['true_w0'])]\n elif measure == 'w0 Bias':\n tmp_y = [BIAS(dataset['first_guessed_w0'], dataset['true_w0'])]\n\n elif measure == 'sfc_h RMSE':\n tmp_y = [RMSE(dataset['first_guess_surface_h'], dataset['true_surface_h'])]\n elif measure == 'sfc_h Diff':\n tmp_y = [DIFF(dataset['first_guess_surface_h'], dataset['true_surface_h'])]\n elif measure == 'sfc_h Bias':\n tmp_y = [DIFF(dataset['first_guess_surface_h'], dataset['true_surface_h'])]\n\n elif measure == 'widths RMSE':\n tmp_y = [RMSE(dataset['first_guess_widths'], dataset['true_widths'])]\n elif measure == 'widths Diff':\n tmp_y = [DIFF(dataset['first_guess_widths'], dataset['true_widths'])]\n elif measure == 'widths Bias':\n tmp_y = [DIFF(dataset['first_guess_widths'], dataset['true_widths'])]\n\n else:\n raise ValueError('Unknown performance measurement!')\n\n for i in dataset.coords['nr_of_iteration'].values[:max_index + 1] - 1:\n tmp_x.append(dataset['computing_time'][i])\n if measure == 'bed_h RMSE':\n tmp_y.append(RMSE(dataset['guessed_bed_h'][i], dataset['true_bed_h']))\n elif measure == 'bed_h Diff':\n tmp_y.append(DIFF(dataset['guessed_bed_h'][i], dataset['true_bed_h']))\n elif measure == 'bed_h Bias':\n tmp_y.append(BIAS(dataset['guessed_bed_h'][i], dataset['true_bed_h']))\n\n elif measure == 'bed_shape RMSE':\n tmp_y.append(RMSE(dataset['guessed_bed_shape'][i], dataset['true_bed_shape']))\n elif measure == 'bed_shape Diff':\n tmp_y.append(DIFF(dataset['guessed_bed_shape'][i], dataset['true_bed_shape']))\n elif measure == 'bed_shape Bias':\n tmp_y.append(BIAS(dataset['guessed_bed_shape'][i], dataset['true_bed_shape']))\n\n elif measure == 'w0 RMSE':\n tmp_y.append(RMSE(dataset['guessed_w0'][i], dataset['true_w0']))\n elif measure == 'w0 Diff':\n tmp_y.append(DIFF(dataset['guessed_w0'][i], dataset['true_w0']))\n elif measure == 'w0 Bias':\n tmp_y.append(BIAS(dataset['guessed_w0'][i], dataset['true_w0']))\n\n elif measure == 'sfc_h RMSE':\n tmp_y.append(RMSE(dataset['surface_h'][i], dataset['true_surface_h']))\n elif measure == 'sfc_h Diff':\n tmp_y.append(DIFF(dataset['surface_h'][i], dataset['true_surface_h']))\n elif measure == 'sfc_h Bias':\n tmp_y.append(BIAS(dataset['surface_h'][i], dataset['true_surface_h']))\n\n elif measure == 'widths RMSE':\n tmp_y.append(RMSE(dataset['widths'][i], dataset['true_widths']))\n elif measure == 'widths Diff':\n tmp_y.append(DIFF(dataset['widths'][i], dataset['true_widths']))\n elif measure == 'widths Bias':\n tmp_y.append(BIAS(dataset['widths'][i], dataset['true_widths']))\n\n else:\n raise ValueError('Unknown performance measurement!')\n else:\n tmp_x = []\n tmp_y = []\n\n all_x.append(tmp_x)\n all_y.append(tmp_y)\n\n colors = [color_1, color_2, color_3, color_4]\n for i, (x, y) in enumerate(zip(all_x, all_y)):\n ax.plot(x, y,\n '.-',\n lw=lw,\n ms=ms,\n c=colors[i])\n\n #ax.legend((),(),title=measure, loc='best')\n #ax.axvline(60, alpha=0.5, c='gray', ls='--')\n # ax.axvline(20, alpha=0.5, c='gray', ls='--')\n #if xlim is not None:\n # ax.set_xlim(xlim)\n ax.tick_params(axis='both', colors=axis_color, width=lw)\n ax.spines['bottom'].set_color(axis_color)\n ax.spines['bottom'].set_linewidth(lw)\n ax.spines['left'].set_color(axis_color)\n ax.spines['left'].set_linewidth(lw)\n ax.spines['top'].set_visible(False)\n ax.spines['right'].set_visible(False)\n \n ax.set_xlabel(r'$T_{cpu}$', fontsize=fontsize, c=axis_color)\n ax.set_ylabel(y_label, fontsize=fontsize, c=axis_color)\n \n if ax_xlim is not None:\n ax.set_xlim(ax_xlim)\n \n if annotation is not None:\n ax.text(annotation_x_position, annotation_y_position,\n annotation,\n horizontalalignment='left',\n verticalalignment='center',\n transform = ax.transAxes)", "_____no_output_____" ] ], [ [ "# Define Colors", "_____no_output_____" ] ], [ [ "colors = sns.color_palette(\"colorblind\")\ncolors", "_____no_output_____" ], [ "axis_color = list(colors[7]) + [1.]\ncolor_1 = list(colors[3]) + [1.]\ncolor_2 = list(colors[0]) + [1.]\ncolor_3 = list(colors[4]) + [1.]\ncolor_4 = list(colors[2]) + [1.]\nglacier_color = glacier_color = list(colors[9]) + [.5]", "_____no_output_____" ] ], [ [ "# Import Data", "_____no_output_____" ] ], [ [ "input_folder = 'plot_data/'\nfilename_scale_1 = 'par_clif_cons_ret_bed_h_and_bed_shape_at_once_scal1reg11.nc'\nfilename_scale_1e4 = 'par_clif_cons_ret_bed_h_and_bed_shape_at_once_scal1e4reg11.nc'\nfilename_separated = 'par_clif_cons_ret_bed_h_and_bed_shape_separatedreg11.nc'\nfilename_calculated = 'par_clif_cons_ret_bed_h_and_bed_shape_calculatedreg11.nc'\n\ndatasets = []\n\nfor filename in [filename_scale_1, filename_separated, filename_scale_1e4, filename_calculated]:\n with xr.open_dataset(input_folder + filename) as ds:\n datasets.append(ds)", "_____no_output_____" ] ], [ [ "# Create figure with performance", "_____no_output_____" ] ], [ [ "#dataset,\nnr_of_iterations = None\nfacecolor = 'white'\nlabels_pad = -700\ncmap = 'Spectral'\nfontsize = 25\nlw=2\nms=10\nannotation_x_position_spatial = -0.15\nannotation_y_position_spatial = 0.9\nannotation_x_position_performance = -0.14\nannotation_y_position_performance = 1.05\n#index_start_first_profil_row = 0\n#index_end_first_profil_row = 6\n#index_start_second_profil_row = 65\n#index_end_second_profil_row = 71\nsave_file = True\nfilename = 'par_methods_overview.pdf'\n\n#plt.rcParams['font.family'] = 'monospace'\nmpl.rcParams.update({'font.size': fontsize})\n\nfig = plt.figure(figsize=(1,1), facecolor='white')\n\n# define grid\ntotal_width = 10\n# define fixed size of spatial subplot\nspatial_height = 2.5\nspatial_y_separation = 0.5\n# define fixed size for performance plot\nperformance_height = 2.5\nperformance_width = 8\nperformance_separation_y = 1\nseparation_y_performance_spatial = 0.5\n# define fixed size for legend\nlegend_height = 3.5\nseparation_x_legend_spatial = 0.5\n\n# fixed size in inch\n# along x axis x-index for locator\nhoriz = [Size.Fixed(total_width), # 0\n ]\n # y-index for locator\nvert = [Size.Fixed(performance_height), # 0 performance row 2\n Size.Fixed(separation_y_performance_spatial),\n Size.Fixed(performance_height), # 2 performance row 1\n Size.Fixed(separation_y_performance_spatial),\n Size.Fixed(spatial_height), # 4 spatial row 2\n Size.Fixed(spatial_y_separation), \n Size.Fixed(spatial_height), # 6 spatial row 1\n Size.Fixed(separation_x_legend_spatial), \n Size.Fixed(legend_height), # 8 legend\n ]\n\n# define indices for subplots for easier changes later\n# spatial heatmap\nspatial_nx = 0\nspatial_nx1 = 1\nspatial_ny_row_1 = 6\nspatial_ny_row_2 = 4\nspatial_annotation = ['(a)', '(b)']\n# performance\nperformance_nx = 0\nperformance_ny_row_1 = 2\nperformance_ny_row_2 = 0\n# legend\nlegend_nx = 0\nlegend_ny = 8\n\n# Position of the grid in the figure\nrect = (0., 0., 1., 1.) \n\n# divide the axes rectangle into grid whose size is specified by horiz * vert\ndivider = Divider(fig, rect, horiz, vert, aspect=False)\n\nwith plt.rc_context({'font.family': 'monospace'}):\n ax = setup_axes(fig, 111)\n im = heatmap(datasets,\n opti_var='bed_h',\n annotation=spatial_annotation[0],\n annotation_x_position=annotation_x_position_spatial,\n annotation_y_position=annotation_y_position_spatial,\n fig=fig,\n ax=ax,\n cmap=cmap,\n grid_color=facecolor,\n presentation=False,\n labels_pad=labels_pad,\n xlim=5,\n nr_of_iterations=nr_of_iterations)\n ax.set_axes_locator(divider.new_locator(nx=spatial_nx,\n nx1=spatial_nx1,\n ny=spatial_ny_row_1))\n \n ax = setup_axes(fig, 111)\n im = heatmap(datasets,\n opti_var='bed_shape',\n annotation=spatial_annotation[1],\n annotation_x_position=annotation_x_position_spatial,\n annotation_y_position=annotation_y_position_spatial,\n fig=fig,\n ax=ax,\n cmap='vlag',\n grid_color=facecolor,\n presentation=False,\n labels_pad=labels_pad,\n xlim=5,\n nr_of_iterations=nr_of_iterations)\n ax.set_axes_locator(divider.new_locator(nx=spatial_nx,\n nx1=spatial_nx1,\n ny=spatial_ny_row_2))\n# add perfomance plot bed_h RMSE\nax = fig.subplots()\nperformance_plot(ax,\n datasets,\n fig=None,\n # 'bed_h RMSE', 'bed_h Diff', 'bed_h Bias',\n # 'bed_shape RMSE', 'bed_shape Diff', 'bed_shape Bias',\n # 'w0 RMSE', 'w0 Diff', 'w0 Bias',\n # 'sfc_h RMSE', 'sfc_h Diff', 'sfc_h Bias',\n # 'widths RMSE', 'widths Diff', 'widths Bias'\n performance_measurement='bed_h RMSE',\n xlim=5,\n y_label='RMSE_b',\n annotation='(c)',\n annotation_x_position=annotation_x_position_performance,\n annotation_y_position=annotation_y_position_performance,\n lw=lw,\n fontsize=fontsize,\n ms=ms,\n nr_of_iterations=nr_of_iterations,\n ax_xlim=[0, 400]\n )\nax.set_axes_locator(divider.new_locator(nx=performance_nx,\n ny=performance_ny_row_1))\n\n# add perfomance plot bed_shape RMSE\nax = fig.subplots()\nperformance_plot(ax,\n datasets,\n fig=None,\n # 'bed_h RMSE', 'bed_h Diff', 'bed_h Bias',\n # 'bed_shape RMSE', 'bed_shape Diff', 'bed_shape Bias',\n # 'w0 RMSE', 'w0 Diff', 'w0 Bias',\n # 'sfc_h RMSE', 'sfc_h Diff', 'sfc_h Bias',\n # 'widths RMSE', 'widths Diff', 'widths Bias'\n performance_measurement='bed_shape RMSE',\n xlim=5,\n y_label='RMSE_Ps',\n annotation='(d)',\n annotation_x_position=annotation_x_position_performance,\n annotation_y_position=annotation_y_position_performance,\n lw=lw,\n fontsize=fontsize,\n ms=ms,\n nr_of_iterations=nr_of_iterations,\n ax_xlim=[0, 350]\n )\nax.set_axes_locator(divider.new_locator(nx=performance_nx,\n ny=performance_ny_row_2))\n\n# add legend\nax = fig.subplots()\nadd_legend2(ax=ax,\n title=(r'$\\bf{cliff}$ with $\\bf{constant}$ width and $\\bf{parabolic}$ shape,' +\n '\\n' +\n r'$\\bf{retreating}$ from an $\\bf{initial~ glacier~ surface}$,' +\n '\\n' + \n r'regularisation parameters $\\lambda_0$ = 1 and $\\lambda_1$ = 100'),\n fontsize=fontsize,\n lw=lw,\n ms=ms,\n labels=['fg: first guess',\n \"and A: 'explicit' without scaling\",\n \"and B: 'iterative'\",\n \"and C: 'explicit' with scaling of 1e-4\",\n \"and D: 'implicit' with no limits\"])\nax.set_axes_locator(divider.new_locator(nx=legend_nx,\n ny=legend_ny))\n\nif save_file:\n fig.savefig(filename, format='pdf', bbox_inches='tight', dpi=300);", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import sqlite3\nfrom pathlib import Path\nimport pandas as pd", "/Users/santiagobasulto/Library/Caches/pypoetry/virtualenvs/pycon-concurrency-tutorial-2020-a5tTVfGc-py3.8/lib/python3.8/site-packages/pandas/compat/__init__.py:117: UserWarning: Could not import the lzma module. Your installed Python is incomplete. Attempting to use lzma compression will result in a RuntimeError.\n warnings.warn(msg)\n" ], [ "SCHEMA = '''\n\ndrop table if exists price;\ncreate table price (\n id integer primary key autoincrement,\n exchange text not null,\n symbol text not null,\n open DECIMAL(10, 4),\n high DECIMAL(10, 4),\n low DECIMAL(10, 4),\n close DECIMAL(10, 4),\n volume DECIMAL(10, 4),\n day DATE not null\n);\n'''", "_____no_output_____" ] ], [ [ "### Prune DB", "_____no_output_____" ] ], [ [ "conn = sqlite3.connect('prices.db')", "_____no_output_____" ], [ "conn.executescript(SCHEMA)", "_____no_output_____" ], [ "conn.commit()", "_____no_output_____" ], [ "conn.close()", "_____no_output_____" ] ], [ [ "### Insert Data", "_____no_output_____" ] ], [ [ "BASE_PATH = Path('crypto_data/')", "_____no_output_____" ], [ "files = list(BASE_PATH.glob('*.csv'))", "_____no_output_____" ], [ "INSERT_STATEMENT = \"\"\"\nINSERT INTO price (\n exchange, symbol, open, high, low, close, volume, day\n) VALUES (?, ?, ?, ?, ?, ?, ?, ?);\n\"\"\"", "_____no_output_____" ], [ "conn = sqlite3.connect('prices.db')", "_____no_output_____" ], [ "for file in files:\n exchange, symbol = file.name[:-4].split('_')\n df = pd.read_csv(str(file))\n df['exchange'] = exchange\n df['symbol'] = symbol\n \n values = df[['exchange', 'symbol', 'OpenPrice', 'HighPrice', 'LowPrice', 'ClosePrice', 'Volume', 'DateTime']].values\n conn.executemany(INSERT_STATEMENT, values)\n conn.commit()", "_____no_output_____" ], [ "conn.close()", "_____no_output_____" ] ], [ [ "### Final Test", "_____no_output_____" ] ], [ [ "conn = sqlite3.connect('prices.db')", "_____no_output_____" ], [ "cursor = conn.cursor()", "_____no_output_____" ], [ "cursor.execute('SELECT COUNT(*) FROM price;')", "_____no_output_____" ], [ "cursor.fetchone()", "_____no_output_____" ], [ "cursor.execute('SELECT * FROM price LIMIT 5;')", "_____no_output_____" ], [ "cursor.fetchall()", "_____no_output_____" ], [ "conn.close()", "_____no_output_____" ] ], [ [ "### Exchanges", "_____no_output_____" ] ], [ [ "conn = sqlite3.connect('prices.db')", "_____no_output_____" ], [ "cursor = conn.cursor()", "_____no_output_____" ], [ "cursor.execute('SELECT DISTINCT exchange FROM price;')", "_____no_output_____" ], [ "cursor.fetchall()", "_____no_output_____" ], [ "cursor.execute('SELECT DISTINCT symbol FROM price;')", "_____no_output_____" ], [ "cursor.fetchall()", "_____no_output_____" ] ], [ [ "### Filtered query:", "_____no_output_____" ] ], [ [ "cursor.execute('SELECT * FROM price WHERE symbol = \"btc\" AND exchange = \"bitfinex\" AND day = \"2019-07-20\";')", "_____no_output_____" ], [ "cursor.fetchall()", "_____no_output_____" ] ] ]

[ "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code" ] ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ [ [ "import numpy as np\nnp.random.rand(15,15).shape", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# import sys\n# #sys.path.insert(0,'../input/dlibpkg/dlib-19.19.0/')\n# sys.path.insert(0,'../input/imutils/imutils-0.5.3/')", "_____no_output_____" ], [ "# !pip install dlib", "_____no_output_____" ], [ "# import dlib\n# from scipy.spatial import distance as dist\n# from imutils.video import FileVideoStream\n# from imutils.video import VideoStream\n# from imutils import face_utils\n# import numpy as np\n# import imutils\n# import time\n# import cv2", "_____no_output_____" ], [ "# def eye_aspect_ratio(eye):\n# # compute the euclidean distances between the two sets of vertical eye landmarks (x, y)-coordinates\n# A = dist.euclidean(eye[1], eye[5])\n# B = dist.euclidean(eye[2], eye[4])\n# # compute the euclidean distance between the horizontal-eye landmark (x, y)-coordinates\n# C = dist.euclidean(eye[0], eye[3])\n# ear = (A + B) / (2.0 * C)\n# return ear\n \n# # define two constants, one for the eye aspect ratio to indicate\n# # blink and then a second constant for the number of consecutive\n# # frames the eye must be below the threshold\n# EYE_AR_THRESH = 0.3\n# EYE_AR_CONSEC_FRAMES = 3\n# # initialize the frame counters and the total number of blinks\n# COUNTER = 0\n# TOTAL = 0\n# print(\"[INFO] loading facial landmark predictor...\")\n# detector = dlib.get_frontal_face_detector()\n# dlib_path = '../input/face-det/shape_predictor_68_face_landmarks.dat'\n# predictor = dlib.shape_predictor(dlib_path)", "_____no_output_____" ], [ "# # loop over frames from the video stream and grab the indexes of the facial landmarks for the left and right eye, respectively\n# (lStart, lEnd) = face_utils.FACIAL_LANDMARKS_IDXS[\"left_eye\"]\n# (rStart, rEnd) = face_utils.FACIAL_LANDMARKS_IDXS[\"right_eye\"]\n# # start the video stream thread\n# print(\"[INFO] starting video stream thread..................\")\n# video_path = '../input/deepfake-detection-challenge/train_sample_videos/dkzvdrzcnr.mp4'\n# vs = FileVideoStream(video_path).start()\n# fileStream = True\n# time.sleep(1.0)\n\n# while True:\n# if fileStream and not vs.more():\n# break\n# frame = vs.read()\n# print(frame.shape)\n# frame = imutils.resize(frame, width=450)\n# print(\"after\")\n# gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)\n\n# # detect faces in the grayscale frame\n# rects = detector(gray, 0)\n# print(rects)\n# # loop over the face detections\n# for rect in rects:\n# # determine the facial landmarks for the face region, then convert the facial landmark (x, y)-coordinates to a NumPy array\n# shape = predictor(gray, rect)\n# shape = face_utils.shape_to_np(shape)\n# # extract the left and right eye coordinates, then use the\n# # coordinates to compute the eye aspect ratio for both eyes\n# leftEye = shape[lStart:lEnd]\n# rightEye = shape[rStart:rEnd]\n# leftEAR = eye_aspect_ratio(leftEye)\n# rightEAR = eye_aspect_ratio(rightEye)\n# # average the eye aspect ratio together for both eyes\n# ear = (leftEAR + rightEAR) / 2.0\n# print(\"ear\")\n# # compute the convex hull for the left and right eye, then\n# # visualize each of the eyes\n# leftEyeHull = cv2.convexHull(leftEye)\n# rightEyeHull = cv2.convexHull(rightEye)\n# cv2.drawContours(frame, [leftEyeHull], -1, (0, 255, 0), 1)\n# cv2.drawContours(frame, [rightEyeHull], -1, (0, 255, 0), 1)\n\n# # check to see if the eye aspect ratio is below the blink\n# # threshold, and if so, increment the blink frame counter\n# if ear < EYE_AR_THRESH:\n# COUNTER += 1\n\n# # otherwise, the eye aspect ratio is not below the blink\n# # threshold\n# else:\n# # if the eyes were closed for a sufficient number of\n# # then increment the total number of blinks\n# if COUNTER >= EYE_AR_CONSEC_FRAMES:\n# TOTAL += 1\n\n# # reset the eye frame counter\n# COUNTER = 0", "_____no_output_____" ], [ "import pandas as pd\nimport numpy as np\nimport matplotlib.pyplot as plt\nimport cv2\nimport os\nfrom tqdm import tqdm,trange\nfrom sklearn.model_selection import train_test_split\nimport sklearn.metrics\n\nimport torch\nimport torch.nn as nn\nimport torch.nn.functional as F\n\nimport warnings\nwarnings.filterwarnings(\"ignore\")", "_____no_output_____" ] ], [ [ "# Setup Data", "_____no_output_____" ] ], [ [ "#----------Training data set\ndf_train0 = pd.read_json('../input/deepfake/metadata0.json')\ndf_train1 = pd.read_json('../input/deepfake/metadata1.json')\ndf_train2 = pd.read_json('../input/deepfake/metadata2.json')\ndf_train3 = pd.read_json('../input/deepfake/metadata3.json')\ndf_train4 = pd.read_json('../input/deepfake/metadata4.json')\ndf_train5 = pd.read_json('../input/deepfake/metadata5.json')\ndf_train6 = pd.read_json('../input/deepfake/metadata6.json')\ndf_train7 = pd.read_json('../input/deepfake/metadata7.json')\ndf_train8 = pd.read_json('../input/deepfake/metadata8.json')\ndf_train9 = pd.read_json('../input/deepfake/metadata9.json')\ndf_train10 = pd.read_json('../input/deepfake/metadata10.json')\ndf_train11 = pd.read_json('../input/deepfake/metadata11.json')\ndf_train12 = pd.read_json('../input/deepfake/metadata12.json')\ndf_train13 = pd.read_json('../input/deepfake/metadata13.json')\ndf_train14 = pd.read_json('../input/deepfake/metadata14.json')\ndf_train15 = pd.read_json('../input/deepfake/metadata15.json')\ndf_train16 = pd.read_json('../input/deepfake/metadata16.json')\ndf_train17 = pd.read_json('../input/deepfake/metadata17.json')\ndf_train18 = pd.read_json('../input/deepfake/metadata18.json')\ndf_train19 = pd.read_json('../input/deepfake/metadata19.json')\ndf_train20 = pd.read_json('../input/deepfake/metadata20.json')\ndf_train21 = pd.read_json('../input/deepfake/metadata21.json')\ndf_train22 = pd.read_json('../input/deepfake/metadata22.json')\ndf_train23 = pd.read_json('../input/deepfake/metadata23.json')\ndf_train24 = pd.read_json('../input/deepfake/metadata24.json')\ndf_train25 = pd.read_json('../input/deepfake/metadata25.json')\ndf_train26 = pd.read_json('../input/deepfake/metadata26.json')\ndf_train27 = pd.read_json('../input/deepfake/metadata27.json')\ndf_train28 = pd.read_json('../input/deepfake/metadata28.json')\ndf_train29 = pd.read_json('../input/deepfake/metadata29.json')\ndf_train30 = pd.read_json('../input/deepfake/metadata30.json')\ndf_train31 = pd.read_json('../input/deepfake/metadata31.json')\ndf_train32 = pd.read_json('../input/deepfake/metadata32.json')\ndf_train33 = pd.read_json('../input/deepfake/metadata33.json')\ndf_train34 = pd.read_json('../input/deepfake/metadata34.json')\ndf_train35 = pd.read_json('../input/deepfake/metadata35.json')\ndf_train36 = pd.read_json('../input/deepfake/metadata36.json')\ndf_train37 = pd.read_json('../input/deepfake/metadata37.json')\ndf_train38 = pd.read_json('../input/deepfake/metadata38.json')\ndf_train39 = pd.read_json('../input/deepfake/metadata39.json')\ndf_train40 = pd.read_json('../input/deepfake/metadata40.json')\ndf_train41 = pd.read_json('../input/deepfake/metadata41.json')\ndf_train42 = pd.read_json('../input/deepfake/metadata42.json')\ndf_train43 = pd.read_json('../input/deepfake/metadata43.json')\ndf_train44 = pd.read_json('../input/deepfake/metadata44.json')\ndf_train45 = pd.read_json('../input/deepfake/metadata45.json')\ndf_train46 = pd.read_json('../input/deepfake/metadata46.json')\ndf_train = [df_train0 ,df_train1, df_train2, df_train3, df_train4,\n df_train5, df_train6, df_train7, df_train8, df_train9,df_train10,\n df_train11, df_train12, df_train13, df_train14, df_train15,df_train16, \n df_train17, df_train18, df_train19, df_train20, df_train21, df_train22, \n df_train23, df_train24, df_train25, df_train26, df_train27, df_train28, \n df_train29, df_train30, df_train31, df_train32, df_train33, df_train34,\n df_train35, df_train36, df_train37, df_train38, df_train39,\n df_train40, df_train41, df_train42, df_train43, df_train44, df_train45,\n df_train46]\ntrain_nums = [\"%.2d\" % i for i in range(len(df_train)+1)]\n#--------------Validation data set\ndf_val1 = pd.read_json('../input/deepfake/metadata47.json')\ndf_val2 = pd.read_json('../input/deepfake/metadata48.json')\ndf_val3 = pd.read_json('../input/deepfake/metadata49.json')\ndf_val = [df_val1, df_val2, df_val3]\nval_nums =['47', '48', '49']", "_____no_output_____" ], [ "# def get_all_paths(df_list,suffixes_list):\n# LABELS = {'REAL':0,'FAKE':1}\n# paths = []\n# labels = []\n# for df,suffix in tqdm(zip(df_list,suffixes_list),total=len(df_list)):\n# images_names = list(df.columns.values)\n# for img_name in images_names:\n# try:\n# paths.append(get_path(img_name,suffix))\n# labels.append(LABELS[df[img_name]['label']])\n# except Exception as err:\n# #print(err)\n# pass\n# return paths,labels", "_____no_output_____" ], [ "def get_orig_fakes(df):\n orig_fakes = {}\n temp = df.T.groupby(['original',df.T.index,]).count()\n for orig,fake in (list(temp.index)):\n fakes = []\n try:#if key exists\n fakes = orig_fakes[orig]\n fakes.append(fake)\n except KeyError as e:\n fakes.append(fake)\n finally:\n orig_fakes[orig] = fakes\n return orig_fakes\n\ndef get_path(img_name,suffix):\n path = '../input/deepfake/DeepFake'+suffix+'/DeepFake'+suffix+'/' + img_name.replace(\".mp4\",\"\")+ '.jpg'\n if not os.path.exists(path):\n raise Exception\n return path\n\ndef get_all_paths(df_list,suffixes_list):\n paths = []\n labels = []\n count = 0\n for df in tqdm(df_list,total=len(df_list)):\n orig_fakes = get_orig_fakes(df)\n for suffix in suffixes_list:\n try:\n for orig,fakes in orig_fakes.items():\n paths.append(get_path(orig,suffix))\n labels.append(0)#processing REAL image\n for img_name in fakes:\n paths.append(get_path(img_name,suffix))\n labels.append(1)#processing FAKES image\n except Exception as err:\n count+=1\n pass\n print(\"Exceptions:\",count)\n return paths,labels", "_____no_output_____" ], [ "%%time\nval_img_paths, val_img_labels = get_all_paths(df_val,val_nums)\ntrain_img_paths, train_img_labels = get_all_paths(df_train,train_nums)\nlen(train_img_paths),len(val_img_paths)", "100%|██████████| 3/3 [00:00<00:00, 24.20it/s]\n 2%|▏ | 1/47 [00:00<00:05, 7.74it/s]" ], [ "#NOT IDEMPOTENT\nval_img_labels = val_img_labels[:500]\nval_img_paths = val_img_paths[:500]\nlen(val_img_paths)", "_____no_output_____" ] ], [ [ "# Dataset", "_____no_output_____" ] ], [ [ "def read_img(path):\n return cv2.cvtColor(cv2.imread(path),cv2.COLOR_BGR2RGB)\nimport random\ndef shuffle(X,y):\n new = []\n for m,n in zip(X,y):\n new.append([m,n])\n random.shuffle(new)\n X,y = [],[]\n for path,label in new:\n X.append(path)\n y.append(label)\n return X,y", "_____no_output_____" ] ], [ [ "## FAKE-->1 REAL-->0", "_____no_output_____" ] ], [ [ "def get_data(train_paths, train_y, val_paths, val_y):\n train_X=[]\n for img in tqdm(train_paths):\n train_X.append(read_img(img))\n val_X=[]\n for img in tqdm(val_paths):\n val_X.append(read_img(img))\n train_X, train_y = shuffle(train_X,train_y)\n val_X, val_y = shuffle(val_X,val_y)\n return train_X, val_X, train_y, val_y", "_____no_output_____" ], [ "'''\ndef get_random_sampling(paths, y, val_paths, val_y):\n real=[]\n fake=[]\n for path,label in zip(paths,y):\n if label==0: \n real.append(path)\n else: \n fake.append(path)\n # fake=random.sample(fake,len(real))\n paths,y=[],[]\n for x in real:\n paths.append(x)\n y.append(0)\n for x in fake:\n paths.append(x)\n y.append(1)\n\n real=[]\n fake=[]\n for m,n in zip(val_paths,val_y):\n if n==0:\n real.append(m)\n else:\n fake.append(m)\n # fake=random.sample(fake,len(real))\n val_paths,val_y=[],[]\n for x in real:\n val_paths.append(x)\n val_y.append(0)\n for x in fake:\n val_paths.append(x)\n val_y.append(1)\n \n #training dataset\n X=[]\n for img in tqdm(paths):\n X.append(read_img(img))\n #validation dataset \n val_X=[]\n for img in tqdm(val_paths):\n val_X.append(read_img(img))\n\n # Balance with ffhq dataset\n ffhq = os.listdir('../input/ffhq-face-data-set/thumbnails128x128')\n X_ = []#used for train and val \n for file in tqdm(ffhq):\n im = read_img(f'../input/ffhq-face-data-set/thumbnails128x128/{file}')\n im = cv2.resize(im, (150,150))\n X_.append(im)\n random.shuffle(X_)\n #Appending REAL images from FFHQ dataset\n for i in range(64773 - 12130):\n X.append(X_[i])\n y.append(0)\n del X_[0:64773 - 12130]\n \n for i in range(6108 - 1258):\n val_X.append(X_[i])\n val_y.append(0)\n X, y = shuffle(X,y)\n val_X, val_y = shuffle(val_X,val_y)\n\n return X, val_X, y, val_y\n '''", "_____no_output_____" ], [ "from torch.utils.data import Dataset, DataLoader\nmean = [0.485, 0.456, 0.406]\nstd = [0.229, 0.224, 0.225]\n\nclass ImageDataset(Dataset):\n def __init__(self, X, y, training=True, transform=None):\n self.X = X\n self.y = y\n self.transform = transform\n self.training = training\n\n def __len__(self):\n return len(self.X)\n \n def __getitem__(self, idx):\n if torch.is_tensor(idx):\n idx = idx.tolist()\n \n img = self.X[idx]\n\n if self.transform is not None:\n res = self.transform(image=img)\n img = res['image']\n \n img = np.rollaxis(img, 2, 0)\n # img = np.array(img).astype(np.float32) / 255.\n\n labels = self.y[idx]\n labels = np.array(labels).astype(np.float32)\n return [img, labels]", "_____no_output_____" ] ], [ [ "# Model", "_____no_output_____" ] ], [ [ "!pip install pytorchcv --quiet", "_____no_output_____" ], [ "from pytorchcv.model_provider import get_model\nmodel = get_model(\"xception\", pretrained=True)\n# model = get_model(\"resnet18\", pretrained=True)\nmodel = nn.Sequential(*list(model.children())[:-1]) # Remove original output layer\nmodel[0].final_block.pool = nn.Sequential(nn.AdaptiveAvgPool2d(1))\n# model[0].final_pool = nn.Sequential(nn.AdaptiveAvgPool2d(1))", "_____no_output_____" ], [ "class Head(torch.nn.Module):\n def __init__(self, in_f, out_f):\n super().__init__()\n\n self.f = nn.Flatten()\n self.l = nn.Linear(in_f, 512)\n self.d = nn.Dropout(0.30)\n self.o = nn.Linear(512, out_f)\n self.b1 = nn.BatchNorm1d(in_f)\n self.b2 = nn.BatchNorm1d(512)\n self.r = nn.ReLU()\n\n def forward(self, x):\n x = self.f(x)\n x = self.b1(x)\n x = self.d(x)\n\n x = self.l(x)\n x = self.r(x)\n x = self.b2(x)\n x = self.d(x)\n\n out = self.o(x)\n return out", "_____no_output_____" ], [ "class FCN(torch.nn.Module):\n def __init__(self, base, in_f):\n super().__init__()\n self.base = base\n self.h1 = Head(in_f, 1)\n \n def forward(self, x):\n x = self.base(x)\n return self.h1(x)\n\nmodel = FCN(model, 2048)\nPATH = './model1.pth'\nmodel.load_state_dict(torch.load(PATH))\nmodel.eval()", "_____no_output_____" ] ], [ [ "# Train Functions", "_____no_output_____" ] ], [ [ "def calculate_loss(preds, targets):\n return F.binary_cross_entropy(F.sigmoid(preds), targets)\n\ndef train_model(epoch, optimizer, scheduler=None, history=None):\n model.train()\n total_loss = 0\n \n t = tqdm(train_loader)\n for i, (img_batch, y_batch) in enumerate(t):\n img_batch = img_batch.cuda().float()\n y_batch = y_batch.cuda().float()\n\n optimizer.zero_grad()#to avoid accumulating gradients\n preds_batch = model(img_batch)\n loss = calculate_loss(preds_batch, y_batch)\n \n total_loss += loss\n t.set_description(f'Epoch {epoch+1}/{n_epochs}, LR: %6f, Loss: %.4f'%(optimizer.state_dict()['param_groups'][0]['lr'],total_loss/(i+1)))\n\n if history is not None:\n history.loc[epoch + i / len(X), 'train_loss'] = loss.data.cpu().numpy()\n history.loc[epoch + i / len(X), 'lr'] = optimizer.state_dict()['param_groups'][0]['lr']\n\n loss.backward()#computing gradients\n optimizer.step()#updating parameters\n \n if scheduler is not None:\n scheduler.step()", "_____no_output_____" ], [ "def evaluate_model(epoch, scheduler=None, history=None):\n model.eval()\n total_loss = 0.0\n pred = []\n target = []\n with torch.no_grad():\n for img_batch, y_batch in val_loader:\n img_batch = img_batch.cuda().float()\n y_batch = y_batch.cuda().float()\n preds_batch = model(img_batch)\n loss = calculate_loss(preds_batch, y_batch)\n total_loss += loss\n pred = [*map(F.sigmoid,preds_batch)]\n target = [*map(lambda i: i.data.cpu(),y_batch)]\n \n\n pred = [p.data.cpu().numpy() for p in pred]\n pred2 = pred\n pred = [np.round(p) for p in pred]\n pred = np.array(pred)\n #calculating accuracy\n acc = sklearn.metrics.recall_score(target, pred, average='macro')\n target = [i.item() for i in target]\n pred2 = np.array(pred2).clip(0.1, 0.9)\n #calculating log-loss after clipping \n log_loss = sklearn.metrics.log_loss(target, pred2)\n\n total_loss /= len(val_loader)\n \n if history is not None:\n history.loc[epoch, 'dev_loss'] = total_loss.cpu().numpy()\n \n if scheduler is not None:\n scheduler.step(total_loss)\n\n print(f'Dev loss: %.4f, Acc: %.6f, log_loss: %.6f'%(total_loss,acc,log_loss))\n \n return total_loss", "_____no_output_____" ] ], [ [ "# Dataloaders", "_____no_output_____" ] ], [ [ "X, val_X, y, val_y = get_data(train_img_paths, train_img_labels,val_img_paths, val_img_labels)\n\nprint('There are '+str(y.count(1))+' fake train samples')\nprint('There are '+str(y.count(0))+' real train samples')\nprint('There are '+str(val_y.count(1))+' fake val samples')\nprint('There are '+str(val_y.count(0))+' real val samples')", "100%|██████████| 2286/2286 [00:02<00:00, 882.95it/s]\n100%|██████████| 149/149 [00:00<00:00, 881.23it/s]" ], [ "import albumentations\nfrom albumentations.augmentations.transforms import ShiftScaleRotate, HorizontalFlip, Normalize, RandomBrightnessContrast, MotionBlur, Blur, GaussNoise, JpegCompression\ntrain_transform = albumentations.Compose([\n ShiftScaleRotate(p=0.3, scale_limit=0.25, border_mode=1, rotate_limit=25),\n HorizontalFlip(p=0.2),\n RandomBrightnessContrast(p=0.3, brightness_limit=0.25, contrast_limit=0.5),\n MotionBlur(p=.2),\n GaussNoise(p=.2),\n JpegCompression(p=.2, quality_lower=50),\n Normalize()\n])\nval_transform = albumentations.Compose([Normalize()])\n\ntrain_dataset = ImageDataset(X, y, transform=train_transform)\nval_dataset = ImageDataset(val_X, val_y, transform=val_transform)", "_____no_output_____" ], [ "nrow, ncol = 5, 6\nfig, axes = plt.subplots(nrow, ncol, figsize=(20, 8))\naxes = axes.flatten()\nfor i, ax in enumerate(axes):\n image, label = train_dataset[i]\n image = np.rollaxis(image, 0, 3)\n image = image*std + mean\n image = np.clip(image, 0., 1.)\n ax.imshow(image)\n ax.set_title(f'label: {label}')", "_____no_output_____" ] ], [ [ "# Train", "_____no_output_____" ] ], [ [ "import gc\n\nhistory = pd.DataFrame()\nhistory2 = pd.DataFrame()\n\ntorch.cuda.empty_cache()\ngc.collect()\n\nbest = 1e10\nn_epochs = 20\nbatch_size = 64#BATCH SIZE CHANGED\n\ntrain_loader = DataLoader(dataset=train_dataset, batch_size=batch_size, shuffle=True, num_workers=4)\nval_loader = DataLoader(dataset=val_dataset, batch_size=batch_size, shuffle=False, num_workers=0)\n\nmodel = model.cuda()\n\noptimizer = torch.optim.AdamW(model.parameters(), lr=0.001)\n\nscheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(optimizer, patience=5, mode='min', factor=0.7, verbose=True, min_lr=1e-5)\n\nfor epoch in range(n_epochs):\n torch.cuda.empty_cache()\n gc.collect()\n train_model(epoch, optimizer, scheduler=None, history=history)\n loss = evaluate_model(epoch, scheduler=scheduler, history=history2)\n if loss < best:\n best = loss\n print(f'Saving best model...')\n torch.save(model.state_dict(), f'model2.pth')", "Epoch 1/20, LR: 0.001000, Loss: 0.7097: 100%|██████████| 36/36 [00:13<00:00, 2.73it/s]\n" ], [ "history2.plot()", "_____no_output_____" ], [ "import torch\nw = torch.rand(5)\nw.requires_grad_()\nprint(w) \ns = w.sum() \nprint(s)\ns.backward()\nprint(w.grad) # tensor([1., 1., 1., 1., 1.])\ns.backward()\nprint(w.grad) # tensor([2., 2., 2., 2., 2.])\ns.backward()\nprint(w.grad) # tensor([3., 3., 3., 3., 3.])\ns.backward()\nprint(w.grad) # tensor([4., 4., 4., 4., 4.])", "_____no_output_____" ] ] ]

[ "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import numpy as np\n\n", "_____no_output_____" ], [ "# Input activation bits and fractions\nia_b = np.array([8, 8, 8, 8])\nia_f = np.array([6, 5, 5, 5])\n\n# Weights bits and fractions\nw_b = np.array([4, 4, 4, 4])\nw_f = np.array([4, 4, 4, 4])\n\n# Output activation bits and fractions\noa_b = np.array([8, 8, 8, 8])\noa_f = np.array([5, 5, 5, 5])\n", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Face Recognition\n\nIn this assignment, you will build a face recognition system. Many of the ideas presented here are from [FaceNet](https://arxiv.org/pdf/1503.03832.pdf). In lecture, we also talked about [DeepFace](https://research.fb.com/wp-content/uploads/2016/11/deepface-closing-the-gap-to-human-level-performance-in-face-verification.pdf). \n\nFace recognition problems commonly fall into two categories: \n\n- **Face Verification** - \"is this the claimed person?\". For example, at some airports, you can pass through customs by letting a system scan your passport and then verifying that you (the person carrying the passport) are the correct person. A mobile phone that unlocks using your face is also using face verification. This is a 1:1 matching problem. \n- **Face Recognition** - \"who is this person?\". For example, the video lecture showed a [face recognition video](https://www.youtube.com/watch?v=wr4rx0Spihs) of Baidu employees entering the office without needing to otherwise identify themselves. This is a 1:K matching problem. \n\nFaceNet learns a neural network that encodes a face image into a vector of 128 numbers. By comparing two such vectors, you can then determine if two pictures are of the same person.\n \n**In this assignment, you will:**\n- Implement the triplet loss function\n- Use a pretrained model to map face images into 128-dimensional encodings\n- Use these encodings to perform face verification and face recognition\n\n#### Channels-first notation\n\n* In this exercise, we will be using a pre-trained model which represents ConvNet activations using a **\"channels first\"** convention, as opposed to the \"channels last\" convention used in lecture and previous programming assignments. \n* In other words, a batch of images will be of shape $(m, n_C, n_H, n_W)$ instead of $(m, n_H, n_W, n_C)$. \n* Both of these conventions have a reasonable amount of traction among open-source implementations; there isn't a uniform standard yet within the deep learning community. ", "_____no_output_____" ], [ "## <font color='darkblue'>Updates</font>\n\n#### If you were working on the notebook before this update...\n* The current notebook is version \"3a\".\n* You can find your original work saved in the notebook with the previous version name (\"v3\") \n* To view the file directory, go to the menu \"File->Open\", and this will open a new tab that shows the file directory.\n\n#### List of updates\n* `triplet_loss`: Additional Hints added.\n* `verify`: Hints added.\n* `who_is_it`: corrected hints given in the comments.\n* Spelling and formatting updates for easier reading.\n", "_____no_output_____" ], [ "#### Load packages\nLet's load the required packages. ", "_____no_output_____" ] ], [ [ "from keras.models import Sequential\nfrom keras.layers import Conv2D, ZeroPadding2D, Activation, Input, concatenate\nfrom keras.models import Model\nfrom keras.layers.normalization import BatchNormalization\nfrom keras.layers.pooling import MaxPooling2D, AveragePooling2D\nfrom keras.layers.merge import Concatenate\nfrom keras.layers.core import Lambda, Flatten, Dense\nfrom keras.initializers import glorot_uniform\nfrom keras.engine.topology import Layer\nfrom keras import backend as K\nK.set_image_data_format('channels_first')\nimport cv2\nimport os\nimport numpy as np\nfrom numpy import genfromtxt\nimport pandas as pd\nimport tensorflow as tf\nfrom fr_utils import *\nfrom inception_blocks_v2 import *\n\n%matplotlib inline\n%load_ext autoreload\n%autoreload 2\n\nnp.set_printoptions(threshold=np.nan)", "Using TensorFlow backend.\n" ] ], [ [ "## 0 - Naive Face Verification\n\nIn Face Verification, you're given two images and you have to determine if they are of the same person. The simplest way to do this is to compare the two images pixel-by-pixel. If the distance between the raw images are less than a chosen threshold, it may be the same person! \n\n<img src=\"images/pixel_comparison.png\" style=\"width:380px;height:150px;\">\n<caption><center> <u> <font color='purple'> **Figure 1** </u></center></caption>", "_____no_output_____" ], [ "* Of course, this algorithm performs really poorly, since the pixel values change dramatically due to variations in lighting, orientation of the person's face, even minor changes in head position, and so on. \n* You'll see that rather than using the raw image, you can learn an encoding, $f(img)$. \n* By using an encoding for each image, an element-wise comparison produces a more accurate judgement as to whether two pictures are of the same person.", "_____no_output_____" ], [ "## 1 - Encoding face images into a 128-dimensional vector \n\n### 1.1 - Using a ConvNet to compute encodings\n\nThe FaceNet model takes a lot of data and a long time to train. So following common practice in applied deep learning, let's load weights that someone else has already trained. The network architecture follows the Inception model from [Szegedy *et al.*](https://arxiv.org/abs/1409.4842). We have provided an inception network implementation. You can look in the file `inception_blocks_v2.py` to see how it is implemented (do so by going to \"File->Open...\" at the top of the Jupyter notebook. This opens the file directory that contains the '.py' file). ", "_____no_output_____" ], [ "The key things you need to know are:\n\n- This network uses 96x96 dimensional RGB images as its input. Specifically, inputs a face image (or batch of $m$ face images) as a tensor of shape $(m, n_C, n_H, n_W) = (m, 3, 96, 96)$ \n- It outputs a matrix of shape $(m, 128)$ that encodes each input face image into a 128-dimensional vector\n\nRun the cell below to create the model for face images.", "_____no_output_____" ] ], [ [ "FRmodel = faceRecoModel(input_shape=(3, 96, 96))", "_____no_output_____" ], [ "print(\"Total Params:\", FRmodel.count_params())", "Total Params: 3743280\n" ] ], [ [ "** Expected Output **\n<table>\n<center>\nTotal Params: 3743280\n</center>\n</table>\n", "_____no_output_____" ], [ "By using a 128-neuron fully connected layer as its last layer, the model ensures that the output is an encoding vector of size 128. You then use the encodings to compare two face images as follows:\n\n<img src=\"images/distance_kiank.png\" style=\"width:680px;height:250px;\">\n<caption><center> <u> <font color='purple'> **Figure 2**: <br> </u> <font color='purple'> By computing the distance between two encodings and thresholding, you can determine if the two pictures represent the same person</center></caption>\n\nSo, an encoding is a good one if: \n- The encodings of two images of the same person are quite similar to each other. \n- The encodings of two images of different persons are very different.\n\nThe triplet loss function formalizes this, and tries to \"push\" the encodings of two images of the same person (Anchor and Positive) closer together, while \"pulling\" the encodings of two images of different persons (Anchor, Negative) further apart. \n\n<img src=\"images/triplet_comparison.png\" style=\"width:280px;height:150px;\">\n<br>\n<caption><center> <u> <font color='purple'> **Figure 3**: <br> </u> <font color='purple'> In the next part, we will call the pictures from left to right: Anchor (A), Positive (P), Negative (N) </center></caption>", "_____no_output_____" ], [ "\n\n### 1.2 - The Triplet Loss\n\nFor an image $x$, we denote its encoding $f(x)$, where $f$ is the function computed by the neural network.\n\n<img src=\"images/f_x.png\" style=\"width:380px;height:150px;\">\n\n<!--\nWe will also add a normalization step at the end of our model so that $\\mid \\mid f(x) \\mid \\mid_2 = 1$ (means the vector of encoding should be of norm 1).\n!-->\n\nTraining will use triplets of images $(A, P, N)$: \n\n- A is an \"Anchor\" image--a picture of a person. \n- P is a \"Positive\" image--a picture of the same person as the Anchor image.\n- N is a \"Negative\" image--a picture of a different person than the Anchor image.\n\nThese triplets are picked from our training dataset. We will write $(A^{(i)}, P^{(i)}, N^{(i)})$ to denote the $i$-th training example. \n\nYou'd like to make sure that an image $A^{(i)}$ of an individual is closer to the Positive $P^{(i)}$ than to the Negative image $N^{(i)}$) by at least a margin $\\alpha$:\n\n$$\\mid \\mid f(A^{(i)}) - f(P^{(i)}) \\mid \\mid_2^2 + \\alpha < \\mid \\mid f(A^{(i)}) - f(N^{(i)}) \\mid \\mid_2^2$$\n\nYou would thus like to minimize the following \"triplet cost\":\n\n$$\\mathcal{J} = \\sum^{m}_{i=1} \\large[ \\small \\underbrace{\\mid \\mid f(A^{(i)}) - f(P^{(i)}) \\mid \\mid_2^2}_\\text{(1)} - \\underbrace{\\mid \\mid f(A^{(i)}) - f(N^{(i)}) \\mid \\mid_2^2}_\\text{(2)} + \\alpha \\large ] \\small_+ \\tag{3}$$\n\nHere, we are using the notation \"$[z]_+$\" to denote $max(z,0)$. \n\nNotes:\n- The term (1) is the squared distance between the anchor \"A\" and the positive \"P\" for a given triplet; you want this to be small. \n- The term (2) is the squared distance between the anchor \"A\" and the negative \"N\" for a given triplet, you want this to be relatively large. It has a minus sign preceding it because minimizing the negative of the term is the same as maximizing that term.\n- $\\alpha$ is called the margin. It is a hyperparameter that you pick manually. We will use $\\alpha = 0.2$. \n\nMost implementations also rescale the encoding vectors to haven L2 norm equal to one (i.e., $\\mid \\mid f(img)\\mid \\mid_2$=1); you won't have to worry about that in this assignment.\n\n**Exercise**: Implement the triplet loss as defined by formula (3). Here are the 4 steps:\n1. Compute the distance between the encodings of \"anchor\" and \"positive\": $\\mid \\mid f(A^{(i)}) - f(P^{(i)}) \\mid \\mid_2^2$\n2. Compute the distance between the encodings of \"anchor\" and \"negative\": $\\mid \\mid f(A^{(i)}) - f(N^{(i)}) \\mid \\mid_2^2$\n3. Compute the formula per training example: $ \\mid \\mid f(A^{(i)}) - f(P^{(i)}) \\mid \\mid_2^2 - \\mid \\mid f(A^{(i)}) - f(N^{(i)}) \\mid \\mid_2^2 + \\alpha$\n3. Compute the full formula by taking the max with zero and summing over the training examples:\n$$\\mathcal{J} = \\sum^{m}_{i=1} \\large[ \\small \\mid \\mid f(A^{(i)}) - f(P^{(i)}) \\mid \\mid_2^2 - \\mid \\mid f(A^{(i)}) - f(N^{(i)}) \\mid \\mid_2^2+ \\alpha \\large ] \\small_+ \\tag{3}$$\n\n#### Hints\n* Useful functions: `tf.reduce_sum()`, `tf.square()`, `tf.subtract()`, `tf.add()`, `tf.maximum()`.\n* For steps 1 and 2, you will sum over the entries of $\\mid \\mid f(A^{(i)}) - f(P^{(i)}) \\mid \\mid_2^2$ and $\\mid \\mid f(A^{(i)}) - f(N^{(i)}) \\mid \\mid_2^2$. \n* For step 4 you will sum over the training examples.\n\n#### Additional Hints\n* Recall that the square of the L2 norm is the sum of the squared differences: $||x - y||_{2}^{2} = \\sum_{i=1}^{N}(x_{i} - y_{i})^{2}$\n* Note that the `anchor`, `positive` and `negative` encodings are of shape `(m,128)`, where m is the number of training examples and 128 is the number of elements used to encode a single example.\n* For steps 1 and 2, you will maintain the number of `m` training examples and sum along the 128 values of each encoding. \n[tf.reduce_sum](https://www.tensorflow.org/api_docs/python/tf/math/reduce_sum) has an `axis` parameter. This chooses along which axis the sums are applied. \n* Note that one way to choose the last axis in a tensor is to use negative indexing (`axis=-1`).\n* In step 4, when summing over training examples, the result will be a single scalar value.\n* For `tf.reduce_sum` to sum across all axes, keep the default value `axis=None`.", "_____no_output_____" ] ], [ [ "# GRADED FUNCTION: triplet_loss\n\ndef triplet_loss(y_true, y_pred, alpha = 0.2):\n \"\"\"\n Implementation of the triplet loss as defined by formula (3)\n \n Arguments:\n y_true -- true labels, required when you define a loss in Keras, you don't need it in this function.\n y_pred -- python list containing three objects:\n anchor -- the encodings for the anchor images, of shape (None, 128)\n positive -- the encodings for the positive images, of shape (None, 128)\n negative -- the encodings for the negative images, of shape (None, 128)\n \n Returns:\n loss -- real number, value of the loss\n \"\"\"\n \n anchor, positive, negative = y_pred[0], y_pred[1], y_pred[2]\n \n ### START CODE HERE ### (≈ 4 lines)\n # Step 1: Compute the (encoding) distance between the anchor and the positive\n pos_dist = tf.reduce_sum(tf.square(tf.subtract(anchor, positive)), axis = -1)\n # Step 2: Compute the (encoding) distance between the anchor and the negative\n neg_dist = tf.reduce_sum(tf.square(tf.subtract(anchor, negative)), axis = -1)\n # Step 3: subtract the two previous distances and add alpha.\n basic_loss = pos_dist - neg_dist + alpha\n # Step 4: Take the maximum of basic_loss and 0.0. Sum over the training examples.\n loss = tf.reduce_sum(tf.maximum(basic_loss, 0.0))\n ### END CODE HERE ###\n \n return loss", "_____no_output_____" ], [ "with tf.Session() as test:\n tf.set_random_seed(1)\n y_true = (None, None, None)\n y_pred = (tf.random_normal([3, 128], mean=6, stddev=0.1, seed = 1),\n tf.random_normal([3, 128], mean=1, stddev=1, seed = 1),\n tf.random_normal([3, 128], mean=3, stddev=4, seed = 1))\n loss = triplet_loss(y_true, y_pred)\n \n print(\"loss = \" + str(loss.eval()))", "loss = 528.143\n" ] ], [ [ "**Expected Output**:\n\n<table>\n <tr>\n <td>\n **loss**\n </td>\n <td>\n 528.143\n </td>\n </tr>\n\n</table>", "_____no_output_____" ], [ "## 2 - Loading the pre-trained model\n\nFaceNet is trained by minimizing the triplet loss. But since training requires a lot of data and a lot of computation, we won't train it from scratch here. Instead, we load a previously trained model. Load a model using the following cell; this might take a couple of minutes to run. ", "_____no_output_____" ] ], [ [ "FRmodel.compile(optimizer = 'adam', loss = triplet_loss, metrics = ['accuracy'])\nload_weights_from_FaceNet(FRmodel)", "_____no_output_____" ] ], [ [ "Here are some examples of distances between the encodings between three individuals:\n\n<img src=\"images/distance_matrix.png\" style=\"width:380px;height:200px;\">\n<br>\n<caption><center> <u> <font color='purple'> **Figure 4**:</u> <br> <font color='purple'> Example of distance outputs between three individuals' encodings</center></caption>\n\nLet's now use this model to perform face verification and face recognition! ", "_____no_output_____" ], [ "## 3 - Applying the model", "_____no_output_____" ], [ "You are building a system for an office building where the building manager would like to offer facial recognition to allow the employees to enter the building.\n\nYou'd like to build a **Face verification** system that gives access to the list of people who live or work there. To get admitted, each person has to swipe an ID card (identification card) to identify themselves at the entrance. The face recognition system then checks that they are who they claim to be.", "_____no_output_____" ], [ "### 3.1 - Face Verification\n\nLet's build a database containing one encoding vector for each person who is allowed to enter the office. To generate the encoding we use `img_to_encoding(image_path, model)`, which runs the forward propagation of the model on the specified image. \n\nRun the following code to build the database (represented as a python dictionary). This database maps each person's name to a 128-dimensional encoding of their face.", "_____no_output_____" ] ], [ [ "database = {}\ndatabase[\"danielle\"] = img_to_encoding(\"images/danielle.png\", FRmodel)\ndatabase[\"younes\"] = img_to_encoding(\"images/younes.jpg\", FRmodel)\ndatabase[\"tian\"] = img_to_encoding(\"images/tian.jpg\", FRmodel)\ndatabase[\"andrew\"] = img_to_encoding(\"images/andrew.jpg\", FRmodel)\ndatabase[\"kian\"] = img_to_encoding(\"images/kian.jpg\", FRmodel)\ndatabase[\"dan\"] = img_to_encoding(\"images/dan.jpg\", FRmodel)\ndatabase[\"sebastiano\"] = img_to_encoding(\"images/sebastiano.jpg\", FRmodel)\ndatabase[\"bertrand\"] = img_to_encoding(\"images/bertrand.jpg\", FRmodel)\ndatabase[\"kevin\"] = img_to_encoding(\"images/kevin.jpg\", FRmodel)\ndatabase[\"felix\"] = img_to_encoding(\"images/felix.jpg\", FRmodel)\ndatabase[\"benoit\"] = img_to_encoding(\"images/benoit.jpg\", FRmodel)\ndatabase[\"arnaud\"] = img_to_encoding(\"images/arnaud.jpg\", FRmodel)", "_____no_output_____" ] ], [ [ "Now, when someone shows up at your front door and swipes their ID card (thus giving you their name), you can look up their encoding in the database, and use it to check if the person standing at the front door matches the name on the ID.\n\n**Exercise**: Implement the verify() function which checks if the front-door camera picture (`image_path`) is actually the person called \"identity\". You will have to go through the following steps:\n1. Compute the encoding of the image from `image_path`.\n2. Compute the distance between this encoding and the encoding of the identity image stored in the database.\n3. Open the door if the distance is less than 0.7, else do not open it.\n\n\n* As presented above, you should use the L2 distance [np.linalg.norm](https://docs.scipy.org/doc/numpy/reference/generated/numpy.linalg.norm.html). \n* (Note: In this implementation, compare the L2 distance, not the square of the L2 distance, to the threshold 0.7.) \n\n#### Hints\n* `identity` is a string that is also a key in the `database` dictionary.\n* `img_to_encoding` has two parameters: the `image_path` and `model`.", "_____no_output_____" ] ], [ [ "# GRADED FUNCTION: verify\n\ndef verify(image_path, identity, database, model):\n \"\"\"\n Function that verifies if the person on the \"image_path\" image is \"identity\".\n \n Arguments:\n image_path -- path to an image\n identity -- string, name of the person you'd like to verify the identity. Has to be an employee who works in the office.\n database -- python dictionary mapping names of allowed people's names (strings) to their encodings (vectors).\n model -- your Inception model instance in Keras\n \n Returns:\n dist -- distance between the image_path and the image of \"identity\" in the database.\n door_open -- True, if the door should open. False otherwise.\n \"\"\"\n \n ### START CODE HERE ###\n \n # Step 1: Compute the encoding for the image. Use img_to_encoding() see example above. (≈ 1 line)\n encoding = img_to_encoding(image_path, model)\n \n # Step 2: Compute distance with identity's image (≈ 1 line)\n dist = np.linalg.norm(encoding - database[identity])\n \n # Step 3: Open the door if dist < 0.7, else don't open (≈ 3 lines)\n if dist < 0.7:\n print(\"It's \" + str(identity) + \", welcome in!\")\n door_open = True\n else:\n print(\"It's not \" + str(identity) + \", please go away\")\n door_open = False\n \n ### END CODE HERE ###\n \n return dist, door_open", "_____no_output_____" ] ], [ [ "Younes is trying to enter the office and the camera takes a picture of him (\"images/camera_0.jpg\"). Let's run your verification algorithm on this picture:\n\n<img src=\"images/camera_0.jpg\" style=\"width:100px;height:100px;\">", "_____no_output_____" ] ], [ [ "verify(\"images/camera_0.jpg\", \"younes\", database, FRmodel)", "It's younes, welcome in!\n" ] ], [ [ "**Expected Output**:\n\n<table>\n <tr>\n <td>\n **It's younes, welcome in!**\n </td>\n <td>\n (0.65939283, True)\n </td>\n </tr>\n\n</table>", "_____no_output_____" ], [ "Benoit, who does not work in the office, stole Kian's ID card and tried to enter the office. The camera took a picture of Benoit (\"images/camera_2.jpg). Let's run the verification algorithm to check if benoit can enter.\n<img src=\"images/camera_2.jpg\" style=\"width:100px;height:100px;\">", "_____no_output_____" ] ], [ [ "verify(\"images/camera_2.jpg\", \"kian\", database, FRmodel)", "It's not kian, please go away\n" ] ], [ [ "**Expected Output**:\n\n<table>\n <tr>\n <td>\n **It's not kian, please go away**\n </td>\n <td>\n (0.86224014, False)\n </td>\n </tr>\n\n</table>", "_____no_output_____" ], [ "### 3.2 - Face Recognition\n\nYour face verification system is mostly working well. But since Kian got his ID card stolen, when he came back to the office the next day and couldn't get in! \n\nTo solve this, you'd like to change your face verification system to a face recognition system. This way, no one has to carry an ID card anymore. An authorized person can just walk up to the building, and the door will unlock for them! \n\nYou'll implement a face recognition system that takes as input an image, and figures out if it is one of the authorized persons (and if so, who). Unlike the previous face verification system, we will no longer get a person's name as one of the inputs. \n\n**Exercise**: Implement `who_is_it()`. You will have to go through the following steps:\n1. Compute the target encoding of the image from image_path\n2. Find the encoding from the database that has smallest distance with the target encoding. \n - Initialize the `min_dist` variable to a large enough number (100). It will help you keep track of what is the closest encoding to the input's encoding.\n - Loop over the database dictionary's names and encodings. To loop use `for (name, db_enc) in database.items()`.\n - Compute the L2 distance between the target \"encoding\" and the current \"encoding\" from the database.\n - If this distance is less than the min_dist, then set `min_dist` to `dist`, and `identity` to `name`.", "_____no_output_____" ] ], [ [ "# GRADED FUNCTION: who_is_it\n\ndef who_is_it(image_path, database, model):\n \"\"\"\n Implements face recognition for the office by finding who is the person on the image_path image.\n \n Arguments:\n image_path -- path to an image\n database -- database containing image encodings along with the name of the person on the image\n model -- your Inception model instance in Keras\n \n Returns:\n min_dist -- the minimum distance between image_path encoding and the encodings from the database\n identity -- string, the name prediction for the person on image_path\n \"\"\"\n \n ### START CODE HERE ### \n \n ## Step 1: Compute the target \"encoding\" for the image. Use img_to_encoding() see example above. ## (≈ 1 line)\n encoding = img_to_encoding(image_path, model)\n \n ## Step 2: Find the closest encoding ##\n \n # Initialize \"min_dist\" to a large value, say 100 (≈1 line)\n min_dist = 100\n \n # Loop over the database dictionary's names and encodings.\n for (name, db_enc) in database.items():\n \n # Compute L2 distance between the target \"encoding\" and the current db_enc from the database. (≈ 1 line)\n dist = np.linalg.norm(encoding - db_enc)\n\n # If this distance is less than the min_dist, then set min_dist to dist, and identity to name. (≈ 3 lines)\n if dist < min_dist:\n min_dist = dist\n identity = name\n\n ### END CODE HERE ###\n \n if min_dist > 0.7:\n print(\"Not in the database.\")\n else:\n print (\"it's \" + str(identity) + \", the distance is \" + str(min_dist))\n \n return min_dist, identity", "_____no_output_____" ] ], [ [ "Younes is at the front-door and the camera takes a picture of him (\"images/camera_0.jpg\"). Let's see if your who_it_is() algorithm identifies Younes. ", "_____no_output_____" ] ], [ [ "who_is_it(\"images/camera_0.jpg\", database, FRmodel)", "it's younes, the distance is 0.659393\n" ] ], [ [ "**Expected Output**:\n\n<table>\n <tr>\n <td>\n **it's younes, the distance is 0.659393**\n </td>\n <td>\n (0.65939283, 'younes')\n </td>\n </tr>\n\n</table>", "_____no_output_____" ], [ "You can change \"`camera_0.jpg`\" (picture of younes) to \"`camera_1.jpg`\" (picture of bertrand) and see the result.", "_____no_output_____" ], [ "#### Congratulations!\n\n* Your face recognition system is working well! It only lets in authorized persons, and people don't need to carry an ID card around anymore! \n* You've now seen how a state-of-the-art face recognition system works.\n\n#### Ways to improve your facial recognition model\nAlthough we won't implement it here, here are some ways to further improve the algorithm:\n- Put more images of each person (under different lighting conditions, taken on different days, etc.) into the database. Then given a new image, compare the new face to multiple pictures of the person. This would increase accuracy.\n- Crop the images to just contain the face, and less of the \"border\" region around the face. This preprocessing removes some of the irrelevant pixels around the face, and also makes the algorithm more robust.\n", "_____no_output_____" ], [ "## Key points to remember\n- Face verification solves an easier 1:1 matching problem; face recognition addresses a harder 1:K matching problem. \n- The triplet loss is an effective loss function for training a neural network to learn an encoding of a face image.\n- The same encoding can be used for verification and recognition. Measuring distances between two images' encodings allows you to determine whether they are pictures of the same person. ", "_____no_output_____" ], [ "Congrats on finishing this assignment! \n", "_____no_output_____" ], [ "### References:\n\n- Florian Schroff, Dmitry Kalenichenko, James Philbin (2015). [FaceNet: A Unified Embedding for Face Recognition and Clustering](https://arxiv.org/pdf/1503.03832.pdf)\n- Yaniv Taigman, Ming Yang, Marc'Aurelio Ranzato, Lior Wolf (2014). [DeepFace: Closing the gap to human-level performance in face verification](https://research.fb.com/wp-content/uploads/2016/11/deepface-closing-the-gap-to-human-level-performance-in-face-verification.pdf) \n- The pretrained model we use is inspired by Victor Sy Wang's implementation and was loaded using his code: https://github.com/iwantooxxoox/Keras-OpenFace.\n- Our implementation also took a lot of inspiration from the official FaceNet github repository: https://github.com/davidsandberg/facenet \n", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "from pyspark import SparkConf, SparkContext\n\nconf = SparkConf().setMaster('local').setAppName('EmailSpam')\nsc = SparkContext(conf=conf)", "_____no_output_____" ], [ "from pyspark.mllib.regression import LabeledPoint\nfrom pyspark.mllib.feature import HashingTF\nfrom pyspark.mllib.classification import LogisticRegressionWithSGD\n\nspam = sc.textFile('spam.txt')\nham = sc.textFile('ham.txt')", "_____no_output_____" ], [ "tf = HashingTF(numFeatures = 10000)\nspamFeatures = spam.map(lambda email: tf.transform(email.split(' ')))\nhamFeatures = ham.map(lambda email: tf.transform(email.split(' ')))", "_____no_output_____" ], [ "# Create LabeledPoint datasets for positive (spam) and negative (ham) examples.\n\npositiveExamples = spamFeatures.map(lambda features: LabeledPoint(1, features))\nnegativeExamples = hamFeatures.map(lambda features: LabeledPoint(0, features))\n\ntrainingData = positiveExamples.union(negativeExamples)\ntrainingData.cache()", "_____no_output_____" ], [ "trainingData.take(2)", "_____no_output_____" ] ], [ [ "### Logistic Regression", "_____no_output_____" ] ], [ [ "model = LogisticRegressionWithSGD.train(trainingData)", "_____no_output_____" ], [ "posTest = tf.transform('O M G GET cheap stuff by sending money to ...'.split(' '))\nnegTest = tf.transform('Hi Dad, I started studying Spark the other ...'.split(' '))\nprint('Prediction for positive test example: %g' % model.predict(posTest))\nprint('Prediction for negative test example: %g' % model.predict(negTest))", "Prediction for positive test example: 0\nPrediction for negative test example: 0\n" ] ], [ [ "### Creating vectors", "_____no_output_____" ] ], [ [ "from numpy import array\nfrom pyspark.mllib.linalg import Vectors\n\ndenseVec1 = array([1.0, 2.0, 3.0])\ndenseVec2 = Vectors.dense([1.0, 2.0, 3.0])\n\nsparseVec1 = Vectors.sparse(4, {0:1.0, 2:2.0})\nsparseVec2 = Vectors.sparse(4, [0, 2], [1.0, 2.0])", "_____no_output_____" ] ], [ [ "### Using HashingTF", "_____no_output_____" ] ], [ [ "sentence = 'hello hello world'\nwords = sentence.split()\ntf = HashingTF(10000)\ntf.transform(words)", "_____no_output_____" ], [ "rdd = sc.wholeTextFiles(\"data\").map(lambda (name, text): text.split())\ntfVectors = tf.transform(rdd)", "_____no_output_____" ] ] ]

[ "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "<a href=\"https://colab.research.google.com/github/DeepInsider/playground-data/blob/master/docs/articles/deeplearningdat.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>", "_____no_output_____" ], [ "##### Copyright 2019 Digital Advantage - Deep Insider.", "_____no_output_____" ] ], [ [ "#@title Licensed under the Apache License, Version 2.0 (the \"License\");\n# you may not use this file except in compliance with the License.\n# You may obtain a copy of the License at\n#\n# https://www.apache.org/licenses/LICENSE-2.0\n#\n# Unless required by applicable law or agreed to in writing, software\n# distributed under the License is distributed on an \"AS IS\" BASIS,\n# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n# See the License for the specific language governing permissions and\n# limitations under the License.", "_____no_output_____" ] ], [ [ "# 連載『機械学習 ＆ ディープラーニング入門（データ構造編）』のノートブック", "_____no_output_____" ], [ "<table valign=\"middle\">\n <td>\n <a target=\"_blank\" href=\"https://deepinsider.jp/tutor/deeplearningdat\"> <img src=\"https://re.deepinsider.jp/img/ml-logo/manabu.svg\"/>Deep Insiderで記事を読む</a>\n </td>\n <td>\n <a target=\"_blank\" href=\"https://colab.research.google.com/github/DeepInsider/playground-data/blob/master/docs/articles/deeplearningdat.ipynb\"> <img src=\"https://re.deepinsider.jp/img/ml-logo/gcolab.svg\" />Google Colabで実行する</a>\n </td>\n <td>\n <a target=\"_blank\" href=\"https://github.com/DeepInsider/playground-data/blob/master/docs/articles/deeplearningdat.ipynb\"> <img src=\"https://re.deepinsider.jp/img/ml-logo/github.svg\" />GitHubでソースコードを見る</a>\n </td>\n</table>", "_____no_output_____" ], [ "※上から順に実行してください。上のコードで実行したものを再利用しているところがあるため、すべて実行しないとエラーになるコードがあります。 \n　すべてのコードを一括実行したい場合は、メニューバーから［ランタイム］－［すべてのセルを実行］をクリックしてください。", "_____no_output_____" ], [ "※このノートブックは「Python 2」でも実行できるようにしていますが、基本的に「Python 3」を利用することをお勧めします。\n　Python 3を利用するには、メニューバーから［ランタイム］－［ランタイムのタイプを変更］を選択すると表示される［ノートブックの設定］ダイアログの、［ランタイムのタイプ］欄で「Python 3」に選択し、その右下にある［保存］ボタンをクリックしてください。", "_____no_output_____" ] ], [ [ "# Python バージョン2への対応\nfrom __future__ import absolute_import\nfrom __future__ import division\nfrom __future__ import print_function\nfrom __future__ import unicode_literals\n\nimport sys\nprint(sys.version_info.major) # 3 # バージョン（メジャー）\nprint(sys.version_info.minor) # 6 # バージョン（マイナー）", "_____no_output_____" ] ], [ [ "## Python言語におけるデータの構造", "_____no_output_____" ], [ "### Pythonにおける「1つの」データの表現", "_____no_output_____" ], [ "#### リスト1-1　「単一の」データを表現するコード", "_____no_output_____" ] ], [ [ "height = 177.2\n\nprint(height) # 177.2と出力される", "_____no_output_____" ] ], [ [ "#### リスト1-2　変数名だけを記述してオブジェクトの評価結果を出力", "_____no_output_____" ] ], [ [ "height # 177.2と出力される", "_____no_output_____" ] ], [ [ "#### リスト1-3　オブジェクト評価結果の出力とprint()関数の出力の違い", "_____no_output_____" ] ], [ [ "import numpy as np\narray2d = np.array([ [ 165.5, 58.4 ],\n [ 177.2, 67.8 ],\n [ 183.2, 83.7 ] ])\n\nprint(array2d) # [[165.5 58.4]\n # [177.2 67.8]\n # [183.2 83.7]]\n\narray2d # array([[165.5, 58.4],\n # [177.2, 67.8],\n # [183.2, 83.7]])", "_____no_output_____" ] ], [ [ "#### リスト2-1　「単一の」データを複数書いて表現するコード", "_____no_output_____" ] ], [ [ "hana_height = 165.5\ntaro_height = 177.2\njiro_height = 183.2\n\nhana_height, taro_height, jiro_height # (165.5, 177.2, 183.2)", "_____no_output_____" ] ], [ [ "#### リスト2-2　「複数（1次元）の」データを表現するコード", "_____no_output_____" ] ], [ [ "heights = [ 165.5, 177.2, 183.2 ]\n\nheights # [165.5, 177.2, 183.2]", "_____no_output_____" ] ], [ [ "### Pythonにおける「複数（2次元）の」データの表現", "_____no_output_____" ], [ "#### リスト3　「複数（2次元）の」データを表現するコード", "_____no_output_____" ] ], [ [ "people = [ [ 165.5, 58.4 ],\n [ 177.2, 67.8 ],\n [ 183.2, 83.7 ] ]\n\npeople # [165.5, 177.2, 183.2]", "_____no_output_____" ] ], [ [ "### Pythonにおける「複数（多次元）の」データの表現", "_____no_output_____" ], [ "#### リスト4　「複数（3次元）の」データを表現するコード", "_____no_output_____" ] ], [ [ "list3d = [\n [ [ 165.5, 58.4 ], [ 177.2, 67.8 ], [ 183.2, 83.7 ] ],\n [ [ 155.5, 48.4 ], [ 167.2, 57.8 ], [ 173.2, 73.7 ] ],\n [ [ 145.5, 38.4 ], [ 157.2, 47.8 ], [ 163.2, 63.7 ] ]\n]\n\nlist3d # [[[165.5, 58.4], [177.2, 67.8], [183.2, 83.7]],\n # [[155.5, 48.4], [167.2, 57.8], [173.2, 73.7]],\n # [[145.5, 38.4], [157.2, 47.8], [163.2, 63.7]]]", "_____no_output_____" ] ], [ [ "## AIプログラムにおけるデータの構造（基本編）", "_____no_output_____" ], [ "### NumPyのインストール", "_____no_output_____" ], [ "#### リスト5-1　`numpy`パッケージをインストールするためのシェルコマンド", "_____no_output_____" ] ], [ [ "!pip install numpy", "_____no_output_____" ] ], [ [ "### numpyモジュールのインポート", "_____no_output_____" ], [ "#### リスト5-2　`numpy`モジュールをインポートするコード例", "_____no_output_____" ] ], [ [ "import numpy as np", "_____no_output_____" ] ], [ [ "### NumPyのデータ型「多次元配列」オブジェクトの作成", "_____no_output_____" ], [ "#### リスト5-3　`array`関数で多次元配列を作成するコード例（値を使用）", "_____no_output_____" ] ], [ [ "array2d = np.array([ [ 165.5, 58.4 ],\n [ 177.2, 67.8 ],\n [ 183.2, 83.7 ] ])\n\narray2d # array([[165.5, 58.4],\n # [177.2, 67.8],\n # [183.2, 83.7]])", "_____no_output_____" ] ], [ [ "#### リスト5-4　`array`関数で多次元配列を作成するコード例（変数を使用）", "_____no_output_____" ] ], [ [ "array3d = np.array(list3d)\n\narray3d # array([[[165.5, 58.4],\n # [177.2, 67.8],\n # [183.2, 83.7]],\n # \n # [[155.5, 48.4],\n # [167.2, 57.8],\n # [173.2, 73.7]],\n # \n # [[145.5, 38.4],\n # [157.2, 47.8],\n # [163.2, 63.7]]])", "_____no_output_____" ] ], [ [ "#### リスト5-5　`ndarray`クラスの`tolist()`メソッドで多次元リストに変換するコード例", "_____no_output_____" ] ], [ [ "tolist3d = array3d.tolist()\n\ntolist3d # [[[165.5, 58.4], [177.2, 67.8], [183.2, 83.7]],\n # [[155.5, 48.4], [167.2, 57.8], [173.2, 73.7]],\n # [[145.5, 38.4], [157.2, 47.8], [163.2, 63.7]]]", "_____no_output_____" ] ], [ [ "## AIプログラムにおけるデータの構造（応用編）", "_____no_output_____" ], [ "### Pandasのインストール", "_____no_output_____" ], [ "#### リスト6　◎pandas◎パッケージをインストールするためのシェルコマンド", "_____no_output_____" ] ], [ [ "!pip install pandas", "_____no_output_____" ] ], [ [ "#### 図7-1　NumPyのデータをPandasで一覧表として表示する例", "_____no_output_____" ] ], [ [ "import pandas as pd\ndf = pd.DataFrame(array2d, columns=['身長', '体重'])\ndf", "_____no_output_____" ] ], [ [ "## AIプログラムにおけるデータの計算", "_____no_output_____" ], [ "### AI・ディープラーニングで数学を使う理由", "_____no_output_____" ], [ "#### リスト7-1　3人の身長の平均を計算するコード例（個別の値を使用）", "_____no_output_____" ] ], [ [ "# hana_height, taro_height, jiro_height = 165.5, 177.2, 183.2 # Lesson 1のリスト2-1で宣言済み\n\naverage_height = (\n hana_height + \n taro_height + \n jiro_height \n) / 3\n\nprint(average_height) # 175.29999999999998", "_____no_output_____" ] ], [ [ "#### リスト7-2　3人の身長と体重の平均を計算するコード例（多次元配列を使用）", "_____no_output_____" ] ], [ [ "import numpy as np\n\narray1d = np.array([ 165.5, 177.2, 183.2 ])\n\naverage_height = np.average(array1d)\n\naverage_height # 175.29999999999998", "_____no_output_____" ] ], [ [ "### NumPyを使った計算", "_____no_output_____" ], [ "#### リスト8-1　3行2列の行列のさまざまな特性を表示するコード例", "_____no_output_____" ] ], [ [ "array2d = np.array([ [ 165.5, 58.4 ],\n [ 177.2, 67.8 ],\n [ 183.2, 83.7 ] ])\n\nprint(array2d.shape) # (3, 2)\nprint(array2d.ndim) # 2\nprint(array2d.size) # 6", "_____no_output_____" ] ], [ [ "#### リスト8-2　NumPyを使った行列計算", "_____no_output_____" ] ], [ [ "diet = np.array([ [ 1.0, 0.0 ],\n [ 0.0, 0.9 ] ])\n\nlose_weights = diet @ array2d.T\n# Python 3.5以降の場合。それ以前のPython 2系などの場合は、以下のmatmul関数を使う必要がある\n#lose_weights = np.matmul(diet, array2d.T)\n\nprint(lose_weights.T) # [[165.5 52.56]\n # [177.2 61.02]\n # [183.2 75.33]]", "_____no_output_____" ] ], [ [ "#### リスト8-3　全要素の平均値を算出（身長／体重別ではない）", "_____no_output_____" ] ], [ [ "averages = np.average(array2d)\n\naverages # 122.63333333333334", "_____no_output_____" ] ], [ [ "#### リスト8-4　身長／体重別の平均値を算出", "_____no_output_____" ] ], [ [ "averages = np.average(array2d, axis=0)\n\naverages # array([175.3 , 69.96666667])", "_____no_output_____" ] ], [ [ "#### リスト8-5　3次元配列データでグループごとの身長／体重別の平均値を算出", "_____no_output_____" ] ], [ [ "array3d = np.array(\n [ [ [ 165.5, 58.4 ], [ 177.2, 67.8 ], [ 183.2, 83.7 ] ],\n [ [ 155.5, 48.4 ], [ 167.2, 57.8 ], [ 173.2, 73.7 ] ],\n [ [ 145.5, 38.4 ], [ 157.2, 47.8 ], [ 163.2, 63.7 ] ] ]\n)\n\navr3d = np.average(array3d, axis=1)\n\nprint(avr3d) # [[175.3 69.96666667]\n # [165.3 59.96666667]\n # [155.3 49.96666667]]", "_____no_output_____" ] ], [ [ "## お疲れさまでした。データ構造の学習は修了です。", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "##### Copyright 2019 The TensorFlow Authors.", "_____no_output_____" ] ], [ [ "#@title Licensed under the Apache License, Version 2.0 (the \"License\");\n# you may not use this file except in compliance with the License.\n# You may obtain a copy of the License at\n#\n# https://www.apache.org/licenses/LICENSE-2.0\n#\n# Unless required by applicable law or agreed to in writing, software\n# distributed under the License is distributed on an \"AS IS\" BASIS,\n# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n# See the License for the specific language governing permissions and\n# limitations under the License.", "_____no_output_____" ] ], [ [ "# Time series forecasting", "_____no_output_____" ], [ "<table class=\"tfo-notebook-buttons\" align=\"left\">\n <td>\n <a target=\"_blank\" href=\"https://www.tensorflow.org/tutorials/structured_data/time_series\"><img src=\"https://www.tensorflow.org/images/tf_logo_32px.png\" />View on TensorFlow.org</a>\n </td>\n <td>\n <a target=\"_blank\" href=\"https://colab.research.google.com/github/tensorflow/docs/blob/master/site/en/tutorials/structured_data/time_series.ipynb\"><img src=\"https://www.tensorflow.org/images/colab_logo_32px.png\" />Run in Google Colab</a>\n </td>\n <td>\n <a target=\"_blank\" href=\"https://github.com/tensorflow/docs/blob/master/site/en/tutorials/structured_data/time_series.ipynb\"><img src=\"https://www.tensorflow.org/images/GitHub-Mark-32px.png\" />View source on GitHub</a>\n </td>\n <td>\n <a href=\"https://storage.googleapis.com/tensorflow_docs/docs/site/en/tutorials/structured_data/time_series.ipynb\"><img src=\"https://www.tensorflow.org/images/download_logo_32px.png\" />Download notebook</a>\n </td>\n</table>", "_____no_output_____" ], [ "This tutorial is an introduction to time series forecasting using TensorFlow. It builds a few different styles of models including Convolutional and Recurrent Neural Networks (CNNs and RNNs).\n\nThis is covered in two main parts, with subsections: \n\n* Forecast for a single timestep:\n * A single feature.\n * All features.\n* Forecast multiple steps:\n * Single-shot: Make the predictions all at once.\n * Autoregressive: Make one prediction at a time and feed the output back to the model.", "_____no_output_____" ], [ "## Setup", "_____no_output_____" ] ], [ [ "import os\nimport datetime\n\nimport IPython\nimport IPython.display\nimport matplotlib as mpl\nimport matplotlib.pyplot as plt\nimport numpy as np\nimport pandas as pd\nimport seaborn as sns\nimport tensorflow as tf\n\nmpl.rcParams['figure.figsize'] = (8, 6)\nmpl.rcParams['axes.grid'] = False", "_____no_output_____" ] ], [ [ "## The weather dataset\nThis tutorial uses a <a href=\"https://www.bgc-jena.mpg.de/wetter/\" class=\"external\">weather time series dataset</a> recorded by the <a href=\"https://www.bgc-jena.mpg.de\" class=\"external\">Max Planck Institute for Biogeochemistry</a>.\n\nThis dataset contains 14 different features such as air temperature, atmospheric pressure, and humidity. These were collected every 10 minutes, beginning in 2003. For efficiency, you will use only the data collected between 2009 and 2016. This section of the dataset was prepared by François Chollet for his book [Deep Learning with Python](https://www.manning.com/books/deep-learning-with-python).", "_____no_output_____" ] ], [ [ "zip_path = tf.keras.utils.get_file(\n origin='https://storage.googleapis.com/tensorflow/tf-keras-datasets/jena_climate_2009_2016.csv.zip',\n fname='jena_climate_2009_2016.csv.zip',\n extract=True)\ncsv_path, _ = os.path.splitext(zip_path)", "_____no_output_____" ] ], [ [ "This tutorial will just deal with **hourly predictions**, so start by sub-sampling the data from 10 minute intervals to 1h:", "_____no_output_____" ] ], [ [ "df = pd.read_csv(csv_path)\n# slice [start:stop:step], starting from index 5 take every 6th record.\ndf = df[5::6]\n\ndate_time = pd.to_datetime(df.pop('Date Time'), format='%d.%m.%Y %H:%M:%S')", "_____no_output_____" ] ], [ [ "Let's take a glance at the data. Here are the first few rows:", "_____no_output_____" ] ], [ [ "df.head()", "_____no_output_____" ] ], [ [ "Here is the evolution of a few features over time. ", "_____no_output_____" ] ], [ [ "plot_cols = ['T (degC)', 'p (mbar)', 'rho (g/m**3)']\nplot_features = df[plot_cols]\nplot_features.index = date_time\n_ = plot_features.plot(subplots=True)\n\nplot_features = df[plot_cols][:480]\nplot_features.index = date_time[:480]\n_ = plot_features.plot(subplots=True)", "_____no_output_____" ] ], [ [ "### Inspect and cleanup", "_____no_output_____" ], [ "Next look at the statistics of the dataset:", "_____no_output_____" ] ], [ [ "df.describe().transpose()", "_____no_output_____" ] ], [ [ "#### Wind velocity", "_____no_output_____" ], [ "One thing that should stand out is the `min` value of the wind velocity, `wv (m/s)` and `max. wv (m/s)` columns. This `-9999` is likely erroneous. There's a separate wind direction column, so the velocity should be `>=0`. Replace it with zeros:\n", "_____no_output_____" ] ], [ [ "wv = df['wv (m/s)']\nbad_wv = wv == -9999.0\nwv[bad_wv] = 0.0\n\nmax_wv = df['max. wv (m/s)']\nbad_max_wv = max_wv == -9999.0\nmax_wv[bad_max_wv] = 0.0\n\n# The above inplace edits are reflected in the DataFrame\ndf['wv (m/s)'].min()", "_____no_output_____" ] ], [ [ "### Feature engineering\n\nBefore diving in to build a model it's important to understand your data, and be sure that you're passing the model appropriately formatted data.", "_____no_output_____" ], [ "#### Wind\nThe last column of the data, `wd (deg)`, gives the wind direction in units of degrees. Angles do not make good model inputs, 360° and 0° should be close to each other, and wrap around smoothly. Direction shouldn't matter if the wind is not blowing. \n\nRight now the distribution of wind data looks like this:", "_____no_output_____" ] ], [ [ "plt.hist2d(df['wd (deg)'], df['wv (m/s)'], bins=(50, 50), vmax=400)\nplt.colorbar()\nplt.xlabel('Wind Direction [deg]')\nplt.ylabel('Wind Velocity [m/s]')", "_____no_output_____" ] ], [ [ "But this will be easier for the model to interpret if you convert the wind direction and velocity columns to a wind **vector**:", "_____no_output_____" ] ], [ [ "wv = df.pop('wv (m/s)')\nmax_wv = df.pop('max. wv (m/s)')\n\n# Convert to radians.\nwd_rad = df.pop('wd (deg)')*np.pi / 180\n\n# Calculate the wind x and y components.\ndf['Wx'] = wv*np.cos(wd_rad)\ndf['Wy'] = wv*np.sin(wd_rad)\n\n# Calculate the max wind x and y components.\ndf['max Wx'] = max_wv*np.cos(wd_rad)\ndf['max Wy'] = max_wv*np.sin(wd_rad)", "_____no_output_____" ] ], [ [ "The distribution of wind vectors is much simpler for the model to correctly interpret.", "_____no_output_____" ] ], [ [ "plt.hist2d(df['Wx'], df['Wy'], bins=(50, 50), vmax=400)\nplt.colorbar()\nplt.xlabel('Wind X [m/s]')\nplt.ylabel('Wind Y [m/s]')\nax = plt.gca()\nax.axis('tight')", "_____no_output_____" ] ], [ [ "#### Time", "_____no_output_____" ], [ "Similarly the `Date Time` column is very useful, but not in this string form. Start by converting it to seconds:", "_____no_output_____" ] ], [ [ "timestamp_s = date_time.map(datetime.datetime.timestamp)", "_____no_output_____" ] ], [ [ "Similar to the wind direction the time in seconds is not a useful model input. Being weather data it has clear daily and yearly periodicity. There are many ways you could deal with periodicity.\n\nA simple approach to convert it to a usable signal is to use `sin` and `cos` to convert the time to clear \"Time of day\" and \"Time of year\" signals:", "_____no_output_____" ] ], [ [ "day = 24*60*60\nyear = (365.2425)*day\n\ndf['Day sin'] = np.sin(timestamp_s * (2 * np.pi / day))\ndf['Day cos'] = np.cos(timestamp_s * (2 * np.pi / day))\ndf['Year sin'] = np.sin(timestamp_s * (2 * np.pi / year))\ndf['Year cos'] = np.cos(timestamp_s * (2 * np.pi / year))", "_____no_output_____" ], [ "plt.plot(np.array(df['Day sin'])[:25])\nplt.plot(np.array(df['Day cos'])[:25])\nplt.xlabel('Time [h]')\nplt.title('Time of day signal')", "_____no_output_____" ] ], [ [ "This gives the model access to the most important frequency features. In this case you knew ahead of time which frequencies were important. \n\nIf you didn't know, you can determine which frequencies are important using an `fft`. To check our assumptions, here is the `tf.signal.rfft` of the temperature over time. Note the obvious peaks at frequencies near `1/year` and `1/day`: ", "_____no_output_____" ] ], [ [ "fft = tf.signal.rfft(df['T (degC)'])\nf_per_dataset = np.arange(0, len(fft))\n\nn_samples_h = len(df['T (degC)'])\nhours_per_year = 24*365.2524\nyears_per_dataset = n_samples_h/(hours_per_year)\n\nf_per_year = f_per_dataset/years_per_dataset\nplt.step(f_per_year, np.abs(fft))\nplt.xscale('log')\nplt.ylim(0, 400000)\nplt.xlim([0.1, max(plt.xlim())])\nplt.xticks([1, 365.2524], labels=['1/Year', '1/day'])\n_ = plt.xlabel('Frequency (log scale)')", "_____no_output_____" ] ], [ [ "### Split the data", "_____no_output_____" ], [ "We'll use a `(70%, 20%, 10%)` split for the training, validation, and test sets. Note the data is **not** being randomly shuffled before splitting. This is for two reasons.\n\n1. It ensures that chopping the data into windows of consecutive samples is still possible.\n2. It ensures that the validation/test results are more realistic, being evaluated on data collected after the model was trained.", "_____no_output_____" ] ], [ [ "column_indices = {name: i for i, name in enumerate(df.columns)}\n\nn = len(df)\ntrain_df = df[0:int(n*0.7)]\nval_df = df[int(n*0.7):int(n*0.9)]\ntest_df = df[int(n*0.9):]\n\nnum_features = df.shape[1]", "_____no_output_____" ] ], [ [ "### Normalize the data\n\nIt is important to scale features before training a neural network. Normalization is a common way of doing this scaling. Subtract the mean and divide by the standard deviation of each feature.", "_____no_output_____" ], [ "The mean and standard deviation should only be computed using the training data so that the models have no access to the values in the validation and test sets.\n\nIt's also arguable that the model shouldn't have access to future values in the training set when training, and that this normalization should be done using moving averages. That's not the focus of this tutorial, and the validation and test sets ensure that you get (somewhat) honest metrics. So in the interest of simplicity this tutorial uses a simple average.", "_____no_output_____" ] ], [ [ "train_mean = train_df.mean()\ntrain_std = train_df.std()\n\ntrain_df = (train_df - train_mean) / train_std\nval_df = (val_df - train_mean) / train_std\ntest_df = (test_df - train_mean) / train_std", "_____no_output_____" ] ], [ [ "Now peek at the distribution of the features. Some features do have long tails, but there are no obvious errors like the `-9999` wind velocity value.", "_____no_output_____" ] ], [ [ "df_std = (df - train_mean) / train_std\ndf_std = df_std.melt(var_name='Column', value_name='Normalized')\nplt.figure(figsize=(12, 6))\nax = sns.violinplot(x='Column', y='Normalized', data=df_std)\n_ = ax.set_xticklabels(df.keys(), rotation=90)", "_____no_output_____" ] ], [ [ "## Data windowing\n\nThe models in this tutorial will make a set of predictions based on a window of consecutive samples from the data. \n\nThe main features of the input windows are:\n\n* The width (number of time steps) of the input and label windows\n* The time offset between them.\n* Which features are used as inputs, labels, or both. \n\nThis tutorial builds a variety of models (including Linear, DNN, CNN and RNN models), and uses them for both:\n\n* *Single-output*, and *multi-output* predictions.\n* *Single-time-step* and *multi-time-step* predictions.\n\nThis section focuses on implementing the data windowing so that it can be reused for all of those models.\n", "_____no_output_____" ], [ "Depending on the task and type of model you may want to generate a variety of data windows. Here are some examples:\n\n1. For example, to make a single prediction 24h into the future, given 24h of history you might define a window like this:\n\n \n\n2. A model that makes a prediction 1h into the future, given 6h of history would need a window like this:\n\n ", "_____no_output_____" ], [ "The rest of this section defines a `WindowGenerator` class. This class can:\n\n1. Handle the indexes and offsets as shown in the diagrams above.\n1. Split windows of features into a `(features, labels)` pairs.\n2. Plot the content of the resulting windows.\n3. Efficiently generate batches of these windows from the training, evaluation, and test data, using `tf.data.Dataset`s.", "_____no_output_____" ], [ "### 1. Indexes and offsets\n\nStart by creating the `WindowGenerator` class. The `__init__` method includes all the necessary logic for the input and label indices.\n\nIt also takes the train, eval, and test dataframes as input. These will be converted to `tf.data.Dataset`s of windows later.", "_____no_output_____" ] ], [ [ "class WindowGenerator():\n def __init__(self, input_width, label_width, shift,\n train_df=train_df, val_df=val_df, test_df=test_df,\n label_columns=None):\n # Store the raw data.\n self.train_df = train_df\n self.val_df = val_df\n self.test_df = test_df\n\n # Work out the label column indices.\n self.label_columns = label_columns\n if label_columns is not None:\n self.label_columns_indices = {name: i for i, name in\n enumerate(label_columns)}\n self.column_indices = {name: i for i, name in\n enumerate(train_df.columns)}\n\n # Work out the window parameters.\n self.input_width = input_width\n self.label_width = label_width\n self.shift = shift\n\n self.total_window_size = input_width + shift\n\n self.input_slice = slice(0, input_width)\n self.input_indices = np.arange(self.total_window_size)[self.input_slice]\n\n self.label_start = self.total_window_size - self.label_width\n self.labels_slice = slice(self.label_start, None)\n self.label_indices = np.arange(self.total_window_size)[self.labels_slice]\n\n def __repr__(self):\n return '\\n'.join([\n f'Total window size: {self.total_window_size}',\n f'Input indices: {self.input_indices}',\n f'Label indices: {self.label_indices}',\n f'Label column name(s): {self.label_columns}'])", "_____no_output_____" ] ], [ [ "Here is code to create the 2 windows shown in the diagrams at the start of this section:", "_____no_output_____" ] ], [ [ "w1 = WindowGenerator(input_width=24, label_width=1, shift=24,\n label_columns=['T (degC)'])\nw1", "_____no_output_____" ], [ "w2 = WindowGenerator(input_width=6, label_width=1, shift=1,\n label_columns=['T (degC)'])\nw2", "_____no_output_____" ] ], [ [ "### 2. Split\nGiven a list consecutive inputs, the `split_window` method will convert them to a window of inputs and a window of labels.\n\nThe example `w2`, above, will be split like this:\n\n\n\nThis diagram doesn't show the `features` axis of the data, but this `split_window` function also handles the `label_columns` so it can be used for both the single output and multi-output examples.", "_____no_output_____" ] ], [ [ "def split_window(self, features):\n inputs = features[:, self.input_slice, :]\n labels = features[:, self.labels_slice, :]\n if self.label_columns is not None:\n labels = tf.stack(\n [labels[:, :, self.column_indices[name]] for name in self.label_columns],\n axis=-1)\n\n # Slicing doesn't preserve static shape information, so set the shapes\n # manually. This way the `tf.data.Datasets` are easier to inspect.\n inputs.set_shape([None, self.input_width, None])\n labels.set_shape([None, self.label_width, None])\n\n return inputs, labels\n\nWindowGenerator.split_window = split_window", "_____no_output_____" ] ], [ [ "Try it out:", "_____no_output_____" ] ], [ [ "# Stack three slices, the length of the total window:\nexample_window = tf.stack([np.array(train_df[:w2.total_window_size]),\n np.array(train_df[100:100+w2.total_window_size]),\n np.array(train_df[200:200+w2.total_window_size])])\n\n\nexample_inputs, example_labels = w2.split_window(example_window)\n\nprint('All shapes are: (batch, time, features)')\nprint(f'Window shape: {example_window.shape}')\nprint(f'Inputs shape: {example_inputs.shape}')\nprint(f'labels shape: {example_labels.shape}')", "_____no_output_____" ] ], [ [ "Typically data in TensorFlow is packed into arrays where the outermost index is across examples (the \"batch\" dimension). The middle indices are the \"time\" or \"space\" (width, height) dimension(s). The innermost indices are the features.\n\nThe code above took a batch of 3, 7-timestep windows, with 19 features at each time step. It split them into a batch of 6-timestep, 19 feature inputs, and a 1-timestep 1-feature label. The label only has one feature because the `WindowGenerator` was initialized with `label_columns=['T (degC)']`. Initially this tutorial will build models that predict single output labels.", "_____no_output_____" ], [ "### 3. Plot\n\nHere is a plot method that allows a simple visualization of the split window:", "_____no_output_____" ] ], [ [ "w2.example = example_inputs, example_labels", "_____no_output_____" ], [ "def plot(self, model=None, plot_col='T (degC)', max_subplots=3):\n inputs, labels = self.example\n plt.figure(figsize=(12, 8))\n plot_col_index = self.column_indices[plot_col]\n max_n = min(max_subplots, len(inputs))\n for n in range(max_n):\n plt.subplot(3, 1, n+1)\n plt.ylabel(f'{plot_col} [normed]')\n plt.plot(self.input_indices, inputs[n, :, plot_col_index],\n label='Inputs', marker='.', zorder=-10)\n\n if self.label_columns:\n label_col_index = self.label_columns_indices.get(plot_col, None)\n else:\n label_col_index = plot_col_index\n\n if label_col_index is None:\n continue\n\n plt.scatter(self.label_indices, labels[n, :, label_col_index],\n edgecolors='k', label='Labels', c='#2ca02c', s=64)\n if model is not None:\n predictions = model(inputs)\n plt.scatter(self.label_indices, predictions[n, :, label_col_index],\n marker='X', edgecolors='k', label='Predictions',\n c='#ff7f0e', s=64)\n\n if n == 0:\n plt.legend()\n\n plt.xlabel('Time [h]')\n\nWindowGenerator.plot = plot", "_____no_output_____" ] ], [ [ "This plot aligns inputs, labels, and (later) predictions based on the time that the item refers to:", "_____no_output_____" ] ], [ [ "w2.plot()", "_____no_output_____" ] ], [ [ "You can plot the other columns, but the example window `w2` configuration only has labels for the `T (degC)` column.", "_____no_output_____" ] ], [ [ "w2.plot(plot_col='p (mbar)')", "_____no_output_____" ] ], [ [ "### 4. Create `tf.data.Dataset`s", "_____no_output_____" ], [ "Finally this `make_dataset` method will take a time series `DataFrame` and convert it to a `tf.data.Dataset` of `(input_window, label_window)` pairs using the `preprocessing.timeseries_dataset_from_array` function.", "_____no_output_____" ] ], [ [ "def make_dataset(self, data):\n data = np.array(data, dtype=np.float32)\n ds = tf.keras.preprocessing.timeseries_dataset_from_array(\n data=data,\n targets=None,\n sequence_length=self.total_window_size,\n sequence_stride=1,\n shuffle=True,\n batch_size=32,)\n\n ds = ds.map(self.split_window)\n\n return ds\n\nWindowGenerator.make_dataset = make_dataset", "_____no_output_____" ] ], [ [ "The `WindowGenerator` object holds training, validation and test data. Add properties for accessing them as `tf.data.Datasets` using the above `make_dataset` method. Also add a standard example batch for easy access and plotting:", "_____no_output_____" ] ], [ [ "@property\ndef train(self):\n return self.make_dataset(self.train_df)\n\n@property\ndef val(self):\n return self.make_dataset(self.val_df)\n\n@property\ndef test(self):\n return self.make_dataset(self.test_df)\n\n@property\ndef example(self):\n \"\"\"Get and cache an example batch of `inputs, labels` for plotting.\"\"\"\n result = getattr(self, '_example', None)\n if result is None:\n # No example batch was found, so get one from the `.train` dataset\n result = next(iter(self.train))\n # And cache it for next time\n self._example = result\n return result\n\nWindowGenerator.train = train\nWindowGenerator.val = val\nWindowGenerator.test = test\nWindowGenerator.example = example", "_____no_output_____" ] ], [ [ "Now the `WindowGenerator` object gives you access to the `tf.data.Dataset` objects, so you can easily iterate over the data.\n\nThe `Dataset.element_spec` property tells you the structure, `dtypes` and shapes of the dataset elements.", "_____no_output_____" ] ], [ [ "# Each element is an (inputs, label) pair\nw2.train.element_spec", "_____no_output_____" ] ], [ [ "Iterating over a `Dataset` yields concrete batches:", "_____no_output_____" ] ], [ [ "for example_inputs, example_labels in w2.train.take(1):\n print(f'Inputs shape (batch, time, features): {example_inputs.shape}')\n print(f'Labels shape (batch, time, features): {example_labels.shape}')", "_____no_output_____" ] ], [ [ "## Single step models\n\nThe simplest model you can build on this sort of data is one that predicts a single feature's value, 1 timestep (1h) in the future based only on the current conditions.\n\nSo start by building models to predict the `T (degC)` value 1h into the future.\n\n\n\nConfigure a `WindowGenerator` object to produce these single-step `(input, label)` pairs:", "_____no_output_____" ] ], [ [ "single_step_window = WindowGenerator(\n input_width=1, label_width=1, shift=1,\n label_columns=['T (degC)'])\nsingle_step_window", "_____no_output_____" ] ], [ [ "The `window` object creates `tf.data.Datasets` from the training, validation, and test sets, allowing you to easily iterate over batches of data.\n", "_____no_output_____" ] ], [ [ "for example_inputs, example_labels in single_step_window.train.take(1):\n print(f'Inputs shape (batch, time, features): {example_inputs.shape}')\n print(f'Labels shape (batch, time, features): {example_labels.shape}')", "_____no_output_____" ] ], [ [ "### Baseline\n\nBefore building a trainable model it would be good to have a performance baseline as a point for comparison with the later more complicated models.\n\nThis first task is to predict temperature 1h in the future given the current value of all features. The current values include the current temperature. \n\nSo start with a model that just returns the current temperature as the prediction, predicting \"No change\". This is a reasonable baseline since temperature changes slowly. Of course, this baseline will work less well if you make a prediction further in the future.\n\n", "_____no_output_____" ] ], [ [ "class Baseline(tf.keras.Model):\n def __init__(self, label_index=None):\n super().__init__()\n self.label_index = label_index\n\n def call(self, inputs):\n if self.label_index is None:\n return inputs\n result = inputs[:, :, self.label_index]\n return result[:, :, tf.newaxis]", "_____no_output_____" ] ], [ [ "Instantiate and evaluate this model:", "_____no_output_____" ] ], [ [ "baseline = Baseline(label_index=column_indices['T (degC)'])\n\nbaseline.compile(loss=tf.losses.MeanSquaredError(),\n metrics=[tf.metrics.MeanAbsoluteError()])\n\nval_performance = {}\nperformance = {}\nval_performance['Baseline'] = baseline.evaluate(single_step_window.val)\nperformance['Baseline'] = baseline.evaluate(single_step_window.test, verbose=0)", "_____no_output_____" ] ], [ [ "That printed some performance metrics, but those don't give you a feeling for how well the model is doing.\n\nThe `WindowGenerator` has a plot method, but the plots won't be very interesting with only a single sample. So, create a wider `WindowGenerator` that generates windows 24h of consecutive inputs and labels at a time. \n\nThe `wide_window` doesn't change the way the model operates. The model still makes predictions 1h into the future based on a single input time step. Here the `time` axis acts like the `batch` axis: Each prediction is made independently with no interaction between time steps.", "_____no_output_____" ] ], [ [ "wide_window = WindowGenerator(\n input_width=24, label_width=24, shift=1,\n label_columns=['T (degC)'])\n\nwide_window", "_____no_output_____" ] ], [ [ "This expanded window can be passed directly to the same `baseline` model without any code changes. This is possible because the inputs and labels have the same number of timesteps, and the baseline just forwards the input to the output:\n\n ", "_____no_output_____" ] ], [ [ "print('Input shape:', wide_window.example[0].shape)\nprint('Output shape:', baseline(wide_window.example[0]).shape)", "_____no_output_____" ] ], [ [ "Plotting the baseline model's predictions you can see that it is simply the labels, shifted right by 1h.", "_____no_output_____" ] ], [ [ "wide_window.plot(baseline)", "_____no_output_____" ] ], [ [ "In the above plots of three examples the single step model is run over the course of 24h. This deserves some explaination:\n\n* The blue \"Inputs\" line shows the input temperature at each time step. The model recieves all features, this plot only shows the temperature.\n* The green \"Labels\" dots show the target prediction value. These dots are shown at the prediction time, not the input time. That is why the range of labels is shifted 1 step relative to the inputs.\n* The orange \"Predictions\" crosses are the model's prediction's for each output time step. If the model were predicting perfectly the predictions would land directly on the \"labels\".", "_____no_output_____" ], [ "### Linear model\n\nThe simplest **trainable** model you can apply to this task is to insert linear transformation between the input and output. In this case the output from a time step only depends on that step:\n\n\n\nA `layers.Dense` with no `activation` set is a linear model. The layer only transforms the last axis of the data from `(batch, time, inputs)` to `(batch, time, units)`, it is applied independently to every item across the `batch` and `time` axes.", "_____no_output_____" ] ], [ [ "linear = tf.keras.Sequential([\n tf.keras.layers.Dense(units=1)\n])", "_____no_output_____" ], [ "print('Input shape:', single_step_window.example[0].shape)\nprint('Output shape:', linear(single_step_window.example[0]).shape)", "_____no_output_____" ] ], [ [ "This tutorial trains many models, so package the training procedure into a function:", "_____no_output_____" ] ], [ [ "MAX_EPOCHS = 20\n\ndef compile_and_fit(model, window, patience=2):\n early_stopping = tf.keras.callbacks.EarlyStopping(monitor='val_loss',\n patience=patience,\n mode='min')\n\n model.compile(loss=tf.losses.MeanSquaredError(),\n optimizer=tf.optimizers.Adam(),\n metrics=[tf.metrics.MeanAbsoluteError()])\n\n history = model.fit(window.train, epochs=MAX_EPOCHS,\n validation_data=window.val,\n callbacks=[early_stopping])\n return history", "_____no_output_____" ] ], [ [ "Train the model and evaluate its performance:", "_____no_output_____" ] ], [ [ "history = compile_and_fit(linear, single_step_window)\n\nval_performance['Linear'] = linear.evaluate(single_step_window.val)\nperformance['Linear'] = linear.evaluate(single_step_window.test, verbose=0)", "_____no_output_____" ] ], [ [ "Like the `baseline` model, the linear model can be called on batches of wide windows. Used this way the model makes a set of independent predictions on consecuitive time steps. The `time` axis acts like another `batch` axis. There are no interactions between the predictions at each time step.\n\n", "_____no_output_____" ] ], [ [ "print('Input shape:', wide_window.example[0].shape)\nprint('Output shape:', baseline(wide_window.example[0]).shape)", "_____no_output_____" ] ], [ [ "Here is the plot of its example predictions on the `wide_window`, note how in many cases the prediction is clearly better than just returning the input temperature, but in a few cases it's worse:", "_____no_output_____" ] ], [ [ "wide_window.plot(linear)", "_____no_output_____" ] ], [ [ "One advantage to linear models is that they're relatively simple to interpret.\nYou can pull out the layer's weights, and see the weight assigned to each input:", "_____no_output_____" ] ], [ [ "plt.bar(x = range(len(train_df.columns)),\n height=linear.layers[0].kernel[:,0].numpy())\naxis = plt.gca()\naxis.set_xticks(range(len(train_df.columns)))\n_ = axis.set_xticklabels(train_df.columns, rotation=90)", "_____no_output_____" ] ], [ [ "Sometimes the model doesn't even place the most weight on the input `T (degC)`. This is one of the risks of random initialization. ", "_____no_output_____" ], [ "### Dense\n\nBefore applying models that actually operate on multiple time-steps, it's worth checking the performance of deeper, more powerful, single input step models.\n\nHere's a model similar to the `linear` model, except it stacks several a few `Dense` layers between the input and the output: ", "_____no_output_____" ] ], [ [ "dense = tf.keras.Sequential([\n tf.keras.layers.Dense(units=64, activation='relu'),\n tf.keras.layers.Dense(units=64, activation='relu'),\n tf.keras.layers.Dense(units=1)\n])\n\nhistory = compile_and_fit(dense, single_step_window)\n\nval_performance['Dense'] = dense.evaluate(single_step_window.val)\nperformance['Dense'] = dense.evaluate(single_step_window.test, verbose=0)", "_____no_output_____" ] ], [ [ "### Multi-step dense\n\nA single-time-step model has no context for the current values of its inputs. It can't see how the input features are changing over time. To address this issue the model needs access to multiple time steps when making predictions:\n\n\n", "_____no_output_____" ], [ "The `baseline`, `linear` and `dense` models handled each time step independently. Here the model will take multiple time steps as input to produce a single output.\n\nCreate a `WindowGenerator` that will produce batches of the 3h of inputs and, 1h of labels:", "_____no_output_____" ], [ "Note that the `Window`'s `shift` parameter is relative to the end of the two windows.\n", "_____no_output_____" ] ], [ [ "CONV_WIDTH = 3\nconv_window = WindowGenerator(\n input_width=CONV_WIDTH,\n label_width=1,\n shift=1,\n label_columns=['T (degC)'])\n\nconv_window", "_____no_output_____" ], [ "conv_window.plot()\nplt.title(\"Given 3h as input, predict 1h into the future.\")", "_____no_output_____" ] ], [ [ "You could train a `dense` model on a multiple-input-step window by adding a `layers.Flatten` as the first layer of the model:", "_____no_output_____" ] ], [ [ "multi_step_dense = tf.keras.Sequential([\n # Shape: (time, features) => (time*features)\n tf.keras.layers.Flatten(),\n tf.keras.layers.Dense(units=32, activation='relu'),\n tf.keras.layers.Dense(units=32, activation='relu'),\n tf.keras.layers.Dense(units=1),\n # Add back the time dimension.\n # Shape: (outputs) => (1, outputs)\n tf.keras.layers.Reshape([1, -1]),\n])", "_____no_output_____" ], [ "print('Input shape:', conv_window.example[0].shape)\nprint('Output shape:', multi_step_dense(conv_window.example[0]).shape)", "_____no_output_____" ], [ "history = compile_and_fit(multi_step_dense, conv_window)\n\nIPython.display.clear_output()\nval_performance['Multi step dense'] = multi_step_dense.evaluate(conv_window.val)\nperformance['Multi step dense'] = multi_step_dense.evaluate(conv_window.test, verbose=0)", "_____no_output_____" ], [ "conv_window.plot(multi_step_dense)", "_____no_output_____" ] ], [ [ "The main down-side of this approach is that the resulting model can only be executed on input windows of exactly this shape. ", "_____no_output_____" ] ], [ [ "print('Input shape:', wide_window.example[0].shape)\ntry:\n print('Output shape:', multi_step_dense(wide_window.example[0]).shape)\nexcept Exception as e:\n print(f'\\n{type(e).__name__}:{e}')", "_____no_output_____" ] ], [ [ "The convolutional models in the next section fix this problem.", "_____no_output_____" ], [ "### Convolution neural network\n \nA convolution layer (`layers.Conv1D`) also takes multiple time steps as input to each prediction.", "_____no_output_____" ], [ "Below is the **same** model as `multi_step_dense`, re-written with a convolution. \n\nNote the changes:\n* The `layers.Flatten` and the first `layers.Dense` are replaced by a `layers.Conv1D`.\n* The `layers.Reshape` is no longer necessary since the convolution keeps the time axis in its output.", "_____no_output_____" ] ], [ [ "conv_model = tf.keras.Sequential([\n tf.keras.layers.Conv1D(filters=32,\n kernel_size=(CONV_WIDTH,),\n activation='relu'),\n tf.keras.layers.Dense(units=32, activation='relu'),\n tf.keras.layers.Dense(units=1),\n])", "_____no_output_____" ] ], [ [ "Run it on an example batch to see that the model produces outputs with the expected shape:", "_____no_output_____" ] ], [ [ "print(\"Conv model on `conv_window`\")\nprint('Input shape:', conv_window.example[0].shape)\nprint('Output shape:', conv_model(conv_window.example[0]).shape)", "_____no_output_____" ] ], [ [ "Train and evaluate it on the ` conv_window` and it should give performance similar to the `multi_step_dense` model.", "_____no_output_____" ] ], [ [ "history = compile_and_fit(conv_model, conv_window)\n\nIPython.display.clear_output()\nval_performance['Conv'] = conv_model.evaluate(conv_window.val)\nperformance['Conv'] = conv_model.evaluate(conv_window.test, verbose=0)", "_____no_output_____" ] ], [ [ "The difference between this `conv_model` and the `multi_step_dense` model is that the `conv_model` can be run on inputs of any length. The convolutional layer is applied to a sliding window of inputs:\n\n\n\nIf you run it on wider input, it produces wider output:", "_____no_output_____" ] ], [ [ "print(\"Wide window\")\nprint('Input shape:', wide_window.example[0].shape)\nprint('Labels shape:', wide_window.example[1].shape)\nprint('Output shape:', conv_model(wide_window.example[0]).shape)", "_____no_output_____" ] ], [ [ "Note that the output is shorter than the input. To make training or plotting work, you need the labels, and prediction to have the same length. So build a `WindowGenerator` to produce wide windows with a few extra input time steps so the label and prediction lengths match: ", "_____no_output_____" ] ], [ [ "LABEL_WIDTH = 24\nINPUT_WIDTH = LABEL_WIDTH + (CONV_WIDTH - 1)\nwide_conv_window = WindowGenerator(\n input_width=INPUT_WIDTH,\n label_width=LABEL_WIDTH,\n shift=1,\n label_columns=['T (degC)'])\n\nwide_conv_window", "_____no_output_____" ], [ "print(\"Wide conv window\")\nprint('Input shape:', wide_conv_window.example[0].shape)\nprint('Labels shape:', wide_conv_window.example[1].shape)\nprint('Output shape:', conv_model(wide_conv_window.example[0]).shape)", "_____no_output_____" ] ], [ [ "Now you can plot the model's predictions on a wider window. Note the 3 input time steps before the first prediction. Every prediction here is based on the 3 preceding timesteps:", "_____no_output_____" ] ], [ [ "wide_conv_window.plot(conv_model)", "_____no_output_____" ] ], [ [ "### Recurrent neural network\n\nA Recurrent Neural Network (RNN) is a type of neural network well-suited to time series data. RNNs process a time series step-by-step, maintaining an internal state from time-step to time-step.\n\nFor more details, read the [text generation tutorial](https://www.tensorflow.org/tutorials/text/text_generation) or the [RNN guide](https://www.tensorflow.org/guide/keras/rnn). \n\nIn this tutorial, you will use an RNN layer called Long Short Term Memory ([LSTM](https://www.tensorflow.org/versions/r2.0/api_docs/python/tf/keras/layers/LSTM)).", "_____no_output_____" ], [ "An important constructor argument for all keras RNN layers is the `return_sequences` argument. This setting can configure the layer in one of two ways.\n\n1. If `False`, the default, the layer only returns the output of the final timestep, giving the model time to warm up its internal state before making a single prediction: \n\n\n\n2. If `True` the layer returns an output for each input. This is useful for:\n * Stacking RNN layers. \n * Training a model on multiple timesteps simultaneously.\n\n", "_____no_output_____" ] ], [ [ "lstm_model = tf.keras.models.Sequential([\n # Shape [batch, time, features] => [batch, time, lstm_units]\n tf.keras.layers.LSTM(32, return_sequences=True),\n # Shape => [batch, time, features]\n tf.keras.layers.Dense(units=1)\n])", "_____no_output_____" ] ], [ [ "With `return_sequences=True` the model can be trained on 24h of data at a time.\n\nNote: This will give a pessimistic view of the model's performance. On the first timestep the model has no access to previous steps, and so can't do any better than the simple `linear` and `dense` models shown earlier.", "_____no_output_____" ] ], [ [ "print('Input shape:', wide_window.example[0].shape)\nprint('Output shape:', lstm_model(wide_window.example[0]).shape)", "_____no_output_____" ], [ "history = compile_and_fit(lstm_model, wide_window)\n\nIPython.display.clear_output()\nval_performance['LSTM'] = lstm_model.evaluate(wide_window.val)\nperformance['LSTM'] = lstm_model.evaluate(wide_window.test, verbose=0)", "_____no_output_____" ], [ "wide_window.plot(lstm_model)", "_____no_output_____" ] ], [ [ "### Performance", "_____no_output_____" ], [ "With this dataset typically each of the models does slightly better than the one before it.", "_____no_output_____" ] ], [ [ "x = np.arange(len(performance))\nwidth = 0.3\nmetric_name = 'mean_absolute_error'\nmetric_index = lstm_model.metrics_names.index('mean_absolute_error')\nval_mae = [v[metric_index] for v in val_performance.values()]\ntest_mae = [v[metric_index] for v in performance.values()]\n\nplt.ylabel('mean_absolute_error [T (degC), normalized]')\nplt.bar(x - 0.17, val_mae, width, label='Validation')\nplt.bar(x + 0.17, test_mae, width, label='Test')\nplt.xticks(ticks=x, labels=performance.keys(),\n rotation=45)\n_ = plt.legend()", "_____no_output_____" ], [ "for name, value in performance.items():\n print(f'{name:12s}: {value[1]:0.4f}')", "_____no_output_____" ] ], [ [ "### Multi-output models\n\nThe models so far all predicted a single output feature, `T (degC)`, for a single time step.\n\nAll of these models can be converted to predict multiple features just by changing the number of units in the output layer and adjusting the training windows to include all features in the `labels`.\n", "_____no_output_____" ] ], [ [ "single_step_window = WindowGenerator(\n # `WindowGenerator` returns all features as labels if you \n # don't set the `label_columns` argument.\n input_width=1, label_width=1, shift=1)\n\nwide_window = WindowGenerator(\n input_width=24, label_width=24, shift=1)\n\nfor example_inputs, example_labels in wide_window.train.take(1):\n print(f'Inputs shape (batch, time, features): {example_inputs.shape}')\n print(f'Labels shape (batch, time, features): {example_labels.shape}')", "_____no_output_____" ] ], [ [ "Note above that the `features` axis of the labels now has the same depth as the inputs, instead of 1.", "_____no_output_____" ], [ "#### Baseline\n\nThe same baseline model can be used here, but this time repeating all features instead of selecting a specific `label_index`.", "_____no_output_____" ] ], [ [ "baseline = Baseline()\nbaseline.compile(loss=tf.losses.MeanSquaredError(),\n metrics=[tf.metrics.MeanAbsoluteError()])", "_____no_output_____" ], [ "val_performance = {}\nperformance = {}\nval_performance['Baseline'] = baseline.evaluate(wide_window.val)\nperformance['Baseline'] = baseline.evaluate(wide_window.test, verbose=0)", "_____no_output_____" ] ], [ [ "#### Dense", "_____no_output_____" ] ], [ [ "dense = tf.keras.Sequential([\n tf.keras.layers.Dense(units=64, activation='relu'),\n tf.keras.layers.Dense(units=64, activation='relu'),\n tf.keras.layers.Dense(units=num_features)\n])", "_____no_output_____" ], [ "history = compile_and_fit(dense, single_step_window)\n\nIPython.display.clear_output()\nval_performance['Dense'] = dense.evaluate(single_step_window.val)\nperformance['Dense'] = dense.evaluate(single_step_window.test, verbose=0)", "_____no_output_____" ] ], [ [ "#### RNN\n", "_____no_output_____" ] ], [ [ "%%time\nwide_window = WindowGenerator(\n input_width=24, label_width=24, shift=1)\n\nlstm_model = tf.keras.models.Sequential([\n # Shape [batch, time, features] => [batch, time, lstm_units]\n tf.keras.layers.LSTM(32, return_sequences=True),\n # Shape => [batch, time, features]\n tf.keras.layers.Dense(units=num_features)\n])\n\nhistory = compile_and_fit(lstm_model, wide_window)\n\nIPython.display.clear_output()\nval_performance['LSTM'] = lstm_model.evaluate( wide_window.val)\nperformance['LSTM'] = lstm_model.evaluate( wide_window.test, verbose=0)\n\nprint()", "_____no_output_____" ] ], [ [ "<a id=\"residual\"></a>\n\n#### Advanced: Residual connections\n\nThe `Baseline` model from earlier took advantage of the fact that the sequence doesn't change drastically from time step to time step. Every model trained in this tutorial so far was randomly initialized, and then had to learn that the output is a a small change from the previous time step.\n\nWhile you can get around this issue with careful initialization, it's simpler to build this into the model structure.\n\nIt's common in time series analysis to build models that instead of predicting the next value, predict how the value will change in the next timestep.\nSimilarly, \"Residual networks\" or \"ResNets\" in deep learning refer to architectures where each layer adds to the model's accumulating result.\n\nThat is how you take advantage of the knowledge that the change should be small.\n\n\n\nEssentially this initializes the model to match the `Baseline`. For this task it helps models converge faster, with slightly better performance.", "_____no_output_____" ], [ "This approach can be used in conjunction with any model discussed in this tutorial. \n\nHere it is being applied to the LSTM model, note the use of the `tf.initializers.zeros` to ensure that the initial predicted changes are small, and don't overpower the residual connection. There are no symmetry-breaking concerns for the gradients here, since the `zeros` are only used on the last layer.", "_____no_output_____" ] ], [ [ "class ResidualWrapper(tf.keras.Model):\n def __init__(self, model):\n super().__init__()\n self.model = model\n\n def call(self, inputs, *args, **kwargs):\n delta = self.model(inputs, *args, **kwargs)\n\n # The prediction for each timestep is the input\n # from the previous time step plus the delta\n # calculated by the model.\n return inputs + delta", "_____no_output_____" ], [ "%%time\nresidual_lstm = ResidualWrapper(\n tf.keras.Sequential([\n tf.keras.layers.LSTM(32, return_sequences=True),\n tf.keras.layers.Dense(\n num_features,\n # The predicted deltas should start small\n # So initialize the output layer with zeros\n kernel_initializer=tf.initializers.zeros)\n]))\n\nhistory = compile_and_fit(residual_lstm, wide_window)\n\nIPython.display.clear_output()\nval_performance['Residual LSTM'] = residual_lstm.evaluate(wide_window.val)\nperformance['Residual LSTM'] = residual_lstm.evaluate(wide_window.test, verbose=0)\nprint()", "_____no_output_____" ] ], [ [ "#### Performance", "_____no_output_____" ], [ "Here is the overall performance for these multi-output models.", "_____no_output_____" ] ], [ [ "x = np.arange(len(performance))\nwidth = 0.3\n\nmetric_name = 'mean_absolute_error'\nmetric_index = lstm_model.metrics_names.index('mean_absolute_error')\nval_mae = [v[metric_index] for v in val_performance.values()]\ntest_mae = [v[metric_index] for v in performance.values()]\n\nplt.bar(x - 0.17, val_mae, width, label='Validation')\nplt.bar(x + 0.17, test_mae, width, label='Test')\nplt.xticks(ticks=x, labels=performance.keys(),\n rotation=45)\nplt.ylabel('MAE (average over all outputs)')\n_ = plt.legend()", "_____no_output_____" ], [ "for name, value in performance.items():\n print(f'{name:15s}: {value[1]:0.4f}')", "_____no_output_____" ] ], [ [ "The above performances are averaged across all model outputs.", "_____no_output_____" ], [ "## Multi-step models\n\nBoth the single-output and multiple-output models in the previous sections made **single time step predictions**, 1h into the future.\n\nThis section looks at how to expand these models to make **multiple time step predictions**.\n\nIn a multi-step prediction, the model needs to learn to predict a range of future values. Thus, unlike a single step model, where only a single future point is predicted, a multi-step model predicts a sequence of the future values.\n\nThere are two rough approaches to this:\n\n1. Single shot predictions where the entire time series is predicted at once.\n2. Autoregressive predictions where the model only makes single step predictions and its output is fed back as its input.\n\nIn this section all the models will predict **all the features across all output time steps**.\n", "_____no_output_____" ], [ "For the multi-step model, the training data again consists of hourly samples. However, here, the models will learn to predict 24h of the future, given 24h of the past.\n\nHere is a `Window` object that generates these slices from the dataset:", "_____no_output_____" ] ], [ [ "OUT_STEPS = 24\nmulti_window = WindowGenerator(input_width=24,\n label_width=OUT_STEPS,\n shift=OUT_STEPS)\n\nmulti_window.plot()\nmulti_window", "_____no_output_____" ] ], [ [ "### Baselines", "_____no_output_____" ], [ "A simple baseline for this task is to repeat the last input time step for the required number of output timesteps:\n\n", "_____no_output_____" ] ], [ [ "class MultiStepLastBaseline(tf.keras.Model):\n def call(self, inputs):\n return tf.tile(inputs[:, -1:, :], [1, OUT_STEPS, 1])\n\nlast_baseline = MultiStepLastBaseline()\nlast_baseline.compile(loss=tf.losses.MeanSquaredError(),\n metrics=[tf.metrics.MeanAbsoluteError()])\n\nmulti_val_performance = {}\nmulti_performance = {}\n\nmulti_val_performance['Last'] = last_baseline.evaluate(multi_window.val)\nmulti_performance['Last'] = last_baseline.evaluate(multi_window.test, verbose=0)\nmulti_window.plot(last_baseline)", "_____no_output_____" ] ], [ [ "Since this task is to predict 24h given 24h another simple approach is to repeat the previous day, assuming tomorrow will be similar:\n\n", "_____no_output_____" ] ], [ [ "class RepeatBaseline(tf.keras.Model):\n def call(self, inputs):\n return inputs\n\nrepeat_baseline = RepeatBaseline()\nrepeat_baseline.compile(loss=tf.losses.MeanSquaredError(),\n metrics=[tf.metrics.MeanAbsoluteError()])\n\nmulti_val_performance['Repeat'] = repeat_baseline.evaluate(multi_window.val)\nmulti_performance['Repeat'] = repeat_baseline.evaluate(multi_window.test, verbose=0)\nmulti_window.plot(repeat_baseline)", "_____no_output_____" ] ], [ [ "### Single-shot models\n\nOne high level approach to this problem is use a \"single-shot\" model, where the model makes the entire sequence prediction in a single step.\n\nThis can be implemented efficiently as a `layers.Dense` with `OUT_STEPS*features` output units. The model just needs to reshape that output to the required `(OUTPUT_STEPS, features)`.", "_____no_output_____" ], [ "#### Linear\n\nA simple linear model based on the last input time step does better than either baseline, but is underpowered. The model needs to predict `OUTPUT_STEPS` time steps, from a single input time step with a linear projection. It can only capture a low-dimensional slice of the behavior, likely based mainly on the time of day and time of year.\n\n", "_____no_output_____" ] ], [ [ "multi_linear_model = tf.keras.Sequential([\n # Take the last time-step.\n # Shape [batch, time, features] => [batch, 1, features]\n tf.keras.layers.Lambda(lambda x: x[:, -1:, :]),\n # Shape => [batch, 1, out_steps*features]\n tf.keras.layers.Dense(OUT_STEPS*num_features,\n kernel_initializer=tf.initializers.zeros),\n # Shape => [batch, out_steps, features]\n tf.keras.layers.Reshape([OUT_STEPS, num_features])\n])\n\nhistory = compile_and_fit(multi_linear_model, multi_window)\n\nIPython.display.clear_output()\nmulti_val_performance['Linear'] = multi_linear_model.evaluate(multi_window.val)\nmulti_performance['Linear'] = multi_linear_model.evaluate(multi_window.test, verbose=0)\nmulti_window.plot(multi_linear_model)", "_____no_output_____" ] ], [ [ "#### Dense\n\nAdding a `layers.Dense` between the input and output gives the linear model more power, but is still only based on a single input timestep.", "_____no_output_____" ] ], [ [ "multi_dense_model = tf.keras.Sequential([\n # Take the last time step.\n # Shape [batch, time, features] => [batch, 1, features]\n tf.keras.layers.Lambda(lambda x: x[:, -1:, :]),\n # Shape => [batch, 1, dense_units]\n tf.keras.layers.Dense(512, activation='relu'),\n # Shape => [batch, out_steps*features]\n tf.keras.layers.Dense(OUT_STEPS*num_features,\n kernel_initializer=tf.initializers.zeros),\n # Shape => [batch, out_steps, features]\n tf.keras.layers.Reshape([OUT_STEPS, num_features])\n])\n\nhistory = compile_and_fit(multi_dense_model, multi_window)\n\nIPython.display.clear_output()\nmulti_val_performance['Dense'] = multi_dense_model.evaluate(multi_window.val)\nmulti_performance['Dense'] = multi_dense_model.evaluate(multi_window.test, verbose=0)\nmulti_window.plot(multi_dense_model)", "_____no_output_____" ] ], [ [ "#### CNN", "_____no_output_____" ], [ "A convolutional model makes predictions based on a fixed-width history, which may lead to better performance than the dense model since it can see how things are changing over time:\n\n", "_____no_output_____" ] ], [ [ "CONV_WIDTH = 3\nmulti_conv_model = tf.keras.Sequential([\n # Shape [batch, time, features] => [batch, CONV_WIDTH, features]\n tf.keras.layers.Lambda(lambda x: x[:, -CONV_WIDTH:, :]),\n # Shape => [batch, 1, conv_units]\n tf.keras.layers.Conv1D(256, activation='relu', kernel_size=(CONV_WIDTH)),\n # Shape => [batch, 1, out_steps*features]\n tf.keras.layers.Dense(OUT_STEPS*num_features,\n kernel_initializer=tf.initializers.zeros),\n # Shape => [batch, out_steps, features]\n tf.keras.layers.Reshape([OUT_STEPS, num_features])\n])\n\nhistory = compile_and_fit(multi_conv_model, multi_window)\n\nIPython.display.clear_output()\n\nmulti_val_performance['Conv'] = multi_conv_model.evaluate(multi_window.val)\nmulti_performance['Conv'] = multi_conv_model.evaluate(multi_window.test, verbose=0)\nmulti_window.plot(multi_conv_model)", "_____no_output_____" ] ], [ [ "#### RNN", "_____no_output_____" ], [ "A recurrent model can learn to use a long history of inputs, if it's relevant to the predictions the model is making. Here the model will accumulate internal state for 24h, before making a single prediction for the next 24h.\n\nIn this single-shot format, the LSTM only needs to produce an output at the last time step, so set `return_sequences=False`.\n\n\n", "_____no_output_____" ] ], [ [ "multi_lstm_model = tf.keras.Sequential([\n # Shape [batch, time, features] => [batch, lstm_units]\n # Adding more `lstm_units` just overfits more quickly.\n tf.keras.layers.LSTM(32, return_sequences=False),\n # Shape => [batch, out_steps*features]\n tf.keras.layers.Dense(OUT_STEPS*num_features,\n kernel_initializer=tf.initializers.zeros),\n # Shape => [batch, out_steps, features]\n tf.keras.layers.Reshape([OUT_STEPS, num_features])\n])\n\nhistory = compile_and_fit(multi_lstm_model, multi_window)\n\nIPython.display.clear_output()\n\nmulti_val_performance['LSTM'] = multi_lstm_model.evaluate(multi_window.val)\nmulti_performance['LSTM'] = multi_lstm_model.evaluate(multi_window.test, verbose=0)\nmulti_window.plot(multi_lstm_model)", "_____no_output_____" ] ], [ [ "### Advanced: Autoregressive model\n\nThe above models all predict the entire output sequence as a in a single step.\n\nIn some cases it may be helpful for the model to decompose this prediction into individual time steps. Then each model's output can be fed back into itself at each step and predictions can be made conditioned on the previous one, like in the classic [Generating Sequences With Recurrent Neural Networks](https://arxiv.org/abs/1308.0850).\n\nOne clear advantage to this style of model is that it can be set up to produce output with a varying length.\n\nYou could take any of single single-step multi-output models trained in the first half of this tutorial and run in an autoregressive feedback loop, but here you'll focus on building a model that's been explicitly trained to do that.\n\n\n", "_____no_output_____" ], [ "#### RNN\n\nThis tutorial only builds an autoregressive RNN model, but this pattern could be applied to any model that was designed to output a single timestep.\n\nThe model will have the same basic form as the single-step `LSTM` models: An `LSTM` followed by a `layers.Dense` that converts the `LSTM` outputs to model predictions.\n\nA `layers.LSTM` is a `layers.LSTMCell` wrapped in the higher level `layers.RNN` that manages the state and sequence results for you (See [Keras RNNs](https://www.tensorflow.org/guide/keras/rnn) for details).\n\nIn this case the model has to manually manage the inputs for each step so it uses `layers.LSTMCell` directly for the lower level, single time step interface.", "_____no_output_____" ] ], [ [ "class FeedBack(tf.keras.Model):\n def __init__(self, units, out_steps):\n super().__init__()\n self.out_steps = out_steps\n self.units = units\n self.lstm_cell = tf.keras.layers.LSTMCell(units)\n # Also wrap the LSTMCell in an RNN to simplify the `warmup` method.\n self.lstm_rnn = tf.keras.layers.RNN(self.lstm_cell, return_state=True)\n self.dense = tf.keras.layers.Dense(num_features)", "_____no_output_____" ], [ "feedback_model = FeedBack(units=32, out_steps=OUT_STEPS)", "_____no_output_____" ] ], [ [ "The first method this model needs is a `warmup` method to initialize its internal state based on the inputs. Once trained this state will capture the relevant parts of the input history. This is equivalent to the single-step `LSTM` model from earlier:", "_____no_output_____" ] ], [ [ "def warmup(self, inputs):\n # inputs.shape => (batch, time, features)\n # x.shape => (batch, lstm_units)\n x, *state = self.lstm_rnn(inputs)\n\n # predictions.shape => (batch, features)\n prediction = self.dense(x)\n return prediction, state\n\nFeedBack.warmup = warmup", "_____no_output_____" ] ], [ [ "This method returns a single time-step prediction, and the internal state of the LSTM:", "_____no_output_____" ] ], [ [ "prediction, state = feedback_model.warmup(multi_window.example[0])\nprediction.shape", "_____no_output_____" ] ], [ [ "With the `RNN`'s state, and an initial prediction you can now continue iterating the model feeding the predictions at each step back as the input.\n\nThe simplest approach to collecting the output predictions is to use a python list, and `tf.stack` after the loop.", "_____no_output_____" ], [ "Note: Stacking a python list like this only works with eager-execution, using `Model.compile(..., run_eagerly=True)` for training, or with a fixed length output. For a dynamic output length you would need to use a `tf.TensorArray` instead of a python list, and `tf.range` instead of the python `range`.", "_____no_output_____" ] ], [ [ "def call(self, inputs, training=None):\n # Use a TensorArray to capture dynamically unrolled outputs.\n predictions = []\n # Initialize the lstm state\n prediction, state = self.warmup(inputs)\n\n # Insert the first prediction\n predictions.append(prediction)\n\n # Run the rest of the prediction steps\n for n in range(1, self.out_steps):\n # Use the last prediction as input.\n x = prediction\n # Execute one lstm step.\n x, state = self.lstm_cell(x, states=state,\n training=training)\n # Convert the lstm output to a prediction.\n prediction = self.dense(x)\n # Add the prediction to the output\n predictions.append(prediction)\n\n # predictions.shape => (time, batch, features)\n predictions = tf.stack(predictions)\n # predictions.shape => (batch, time, features)\n predictions = tf.transpose(predictions, [1, 0, 2])\n return predictions\n\nFeedBack.call = call", "_____no_output_____" ] ], [ [ "Test run this model on the example inputs:", "_____no_output_____" ] ], [ [ "print('Output shape (batch, time, features): ', feedback_model(multi_window.example[0]).shape)", "_____no_output_____" ] ], [ [ "Now train the model:", "_____no_output_____" ] ], [ [ "history = compile_and_fit(feedback_model, multi_window)\n\nIPython.display.clear_output()\n\nmulti_val_performance['AR LSTM'] = feedback_model.evaluate(multi_window.val)\nmulti_performance['AR LSTM'] = feedback_model.evaluate(multi_window.test, verbose=0)\nmulti_window.plot(feedback_model)", "_____no_output_____" ] ], [ [ "### Performance", "_____no_output_____" ], [ "There are clearly diminishing returns as a function of model complexity on this problem.", "_____no_output_____" ] ], [ [ "x = np.arange(len(multi_performance))\nwidth = 0.3\n\n\nmetric_name = 'mean_absolute_error'\nmetric_index = lstm_model.metrics_names.index('mean_absolute_error')\nval_mae = [v[metric_index] for v in multi_val_performance.values()]\ntest_mae = [v[metric_index] for v in multi_performance.values()]\n\nplt.bar(x - 0.17, val_mae, width, label='Validation')\nplt.bar(x + 0.17, test_mae, width, label='Test')\nplt.xticks(ticks=x, labels=multi_performance.keys(),\n rotation=45)\nplt.ylabel(f'MAE (average over all times and outputs)')\n_ = plt.legend()", "_____no_output_____" ] ], [ [ "The metrics for the multi-output models in the first half of this tutorial show the performance averaged across all output features. These performances similar but also averaged across output timesteps. ", "_____no_output_____" ] ], [ [ "for name, value in multi_performance.items():\n print(f'{name:8s}: {value[1]:0.4f}')", "_____no_output_____" ] ], [ [ "The gains achieved going from a dense model to convolutional and recurrent models are only a few percent (if any), and the autoregressive model performed clearly worse. So these more complex approaches may not be worth while on **this** problem, but there was no way to know without trying, and these models could be helpful for **your** problem.", "_____no_output_____" ], [ "## Next steps\n\nThis tutorial was a quick introduction to time series forecasting using TensorFlow.\n\n* For further understanding, see:\n * Chapter 15 of [Hands-on Machine Learning with Scikit-Learn, Keras, and TensorFlow](https://www.oreilly.com/library/view/hands-on-machine-learning/9781492032632/), 2nd Edition \n * Chapter 6 of [Deep Learning with Python](https://www.manning.com/books/deep-learning-with-python).\n * Lesson 8 of [Udacity's intro to TensorFlow for deep learning](https://www.udacity.com/course/intro-to-tensorflow-for-deep-learning--ud187), and the [exercise notebooks](https://github.com/tensorflow/examples/tree/master/courses/udacity_intro_to_tensorflow_for_deep_learning) \n* Also remember that you can implement any [classical time series model](https://otexts.com/fpp2/index.html) in TensorFlow, this tutorial just focuses on TensorFlow's built-in functionality.", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Word Embeddings in MySQL\n\nThis example uses the official MySQL Connector within Python3 to store and retrieve various amounts of Word Embeddings.\n\nWe will use a local MySQL database running as a Docker Container for testing purposes. To start the database run:\n\n```\ndocker run -ti --rm --name ohmysql -e MYSQL_ROOT_PASSWORD=mikolov -e MYSQL_DATABASE=embeddings -p 3306:3306 mysql:5.7\n```", "_____no_output_____" ] ], [ [ "import mysql.connector\nimport io\nimport time\nimport numpy\nimport plotly\nfrom tqdm import tqdm_notebook as tqdm", "_____no_output_____" ] ], [ [ "# Dummy Embeddings\n\nFor testing purposes we will use randomly generated numpy arrays as dummy embbeddings.", "_____no_output_____" ] ], [ [ "def embeddings(n=1000, dim=300):\n \"\"\"\n Yield n tuples of random numpy arrays of *dim* length indexed by *n*\n \"\"\"\n idx = 0\n while idx < n:\n yield (str(idx), numpy.random.rand(dim))\n idx += 1", "_____no_output_____" ] ], [ [ "# Conversion Functions\n\nSince we can't just save a NumPy array into the database, we will convert it into a BLOB.", "_____no_output_____" ] ], [ [ "def adapt_array(array):\n \"\"\"\n Using the numpy.save function to save a binary version of the array,\n and BytesIO to catch the stream of data and convert it into a BLOB.\n \"\"\"\n out = io.BytesIO()\n numpy.save(out, array)\n out.seek(0)\n\n return out.read()\n\ndef convert_array(blob):\n \"\"\"\n Using BytesIO to convert the binary version of the array back into a numpy array.\n \"\"\"\n out = io.BytesIO(blob)\n out.seek(0)\n\n return numpy.load(out)", "_____no_output_____" ], [ "connection = mysql.connector.connect(user='root', password='mikolov',\n host='127.0.0.1',\n database='embeddings')\n\ncursor = connection.cursor()\ncursor.execute('CREATE TABLE IF NOT EXISTS `embeddings` (`key` TEXT, `embedding` BLOB);')\nconnection.commit()", "_____no_output_____" ], [ "%%time\nfor key, emb in embeddings():\n arr = adapt_array(emb)\n cursor.execute('INSERT INTO `embeddings` (`key`, `embedding`) VALUES (%s, %s);', (key, arr))\n connection.commit()", "CPU times: user 1.58 s, sys: 170 ms, total: 1.75 s\nWall time: 1min 55s\n" ], [ "%%time\nfor key, _ in embeddings():\n cursor.execute('SELECT embedding FROM `embeddings` WHERE `key`=%s;', (key,))\n data = cursor.fetchone()\n arr = convert_array(data[0])", "CPU times: user 393 ms, sys: 42.7 ms, total: 435 ms\nWall time: 1.26 s\n" ] ], [ [ "# Sample some data\n\nTo test the I/O we will write and read some data from the database. This may take a while.", "_____no_output_____" ] ], [ [ "write_times = []\nread_times = []\ncounts = [500, 1000, 2000, 3000, 4000, 5000]\n\nfor c in counts:\n print(c)\n cursor.execute('DROP TABLE IF EXISTS `embeddings`;')\n cursor.execute('CREATE TABLE IF NOT EXISTS `embeddings` (`key` TEXT, `embedding` BLOB);')\n connection.commit()\n\n start_time_write = time.time()\n for key, emb in tqdm(embeddings(c), total=c):\n arr = adapt_array(emb)\n cursor.execute('INSERT INTO `embeddings` (`key`, `embedding`) VALUES (%s, %s);', (key, arr))\n connection.commit()\n write_times.append(time.time() - start_time_write)\n \n start_time_read = time.time()\n for key, emb in embeddings(c):\n cursor.execute('SELECT embedding FROM `embeddings` WHERE `key`=%s;', (key,))\n data = cursor.fetchone()\n arr = convert_array(data[0])\n read_times.append(time.time() - start_time_read)\n \nprint('DONE')", "500\n" ] ], [ [ "# Results", "_____no_output_____" ] ], [ [ "plotly.offline.init_notebook_mode(connected=True)\ntrace = plotly.graph_objs.Scatter(\n y = write_times,\n x = counts,\n mode = 'lines+markers'\n)\nlayout = plotly.graph_objs.Layout(title=\"MySQL Write Times\",\n yaxis=dict(title='Time in Seconds'),\n xaxis=dict(title='Embedding Count'))\ndata = [trace]\nfig = plotly.graph_objs.Figure(data=data, layout=layout)\nplotly.offline.iplot(fig, filename='jupyter-scatter-write')", "_____no_output_____" ], [ "plotly.offline.init_notebook_mode(connected=True)\ntrace = plotly.graph_objs.Scatter(\n y = read_times,\n x = counts,\n mode = 'lines+markers'\n)\nlayout = plotly.graph_objs.Layout(title=\"MySQL Read Times\",\n yaxis=dict(title='Time in Seconds'),\n xaxis=dict(title='Embedding Count'))\ndata = [trace]\nfig = plotly.graph_objs.Figure(data=data, layout=layout)\nplotly.offline.iplot(fig, filename='jupyter-scatter-read')", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "from future.builtins import next\nimport os\nimport csv\nimport re\nimport logging\nimport optparse\n\nimport dedupe\nfrom unidecode import unidecode\n\nimport pandas as pd", "_____no_output_____" ], [ "pd.options.display.float_format = '{:20,.2f}'.format\npd.set_option('display.max_rows', 5000)\npd.set_option('display.max_columns', 5000)\npd.set_option('display.width', 1000)\npd.set_option('display.max_colwidth', -1)", "_____no_output_____" ], [ "amazon_walmart_all_path = (r'/home/ubuntu/jupyter/ServerX/1_Standard Data Integration/Sample Datasets'\n r'/Processed Data/product_samples/amazon_walmart_all.csv')", "_____no_output_____" ] ], [ [ "## Prepare df and dict corpus", "_____no_output_____" ] ], [ [ "fields_of_interest = [\n 'Id',\n 'name',\n 'producer',\n 'description',\n 'price',\n 'category',\n 'source'\n]", "_____no_output_____" ], [ "amazon_walmart_all_df = pd.read_csv(amazon_walmart_all_path, sep=',', quotechar='\"')[fields_of_interest]", "_____no_output_____" ], [ "amazon_walmart_all_df.dtypes", "_____no_output_____" ], [ "x = amazon_walmart_all_df[amazon_walmart_all_df['category'].isnull()]\nx.head(1)", "_____no_output_____" ], [ "z = amazon_walmart_all_df[amazon_walmart_all_df['producer'].isnull()]\nz.head(1)", "_____no_output_____" ], [ "y = amazon_walmart_all_df[amazon_walmart_all_df['price'].isnull()]\ny.head(1)", "_____no_output_____" ], [ "h = amazon_walmart_all_df[amazon_walmart_all_df['description'].isnull()]\nh.head()", "_____no_output_____" ], [ "amazon_walmart_all_df[amazon_walmart_all_df['name'].isnull()]", "_____no_output_____" ], [ "description_corpus = amazon_walmart_all_df['description'].to_list()\ndescription_corpus = [x for x in description_corpus if str(x) != 'nan']", "_____no_output_____" ], [ "description_corpus[1]", "_____no_output_____" ], [ "category_corpus = amazon_walmart_all_df.drop_duplicates().to_dict('records')", "_____no_output_____" ], [ "categories = list(amazon_walmart_all_df['category'].unique())\ncategories = [x for x in categories if str(x) != 'nan']", "_____no_output_____" ], [ "producer_corpus = amazon_walmart_all_df.drop_duplicates().to_dict('records')", "_____no_output_____" ], [ "producers = list(amazon_walmart_all_df['producer'].unique())\nproducers = [x for x in producers if str(x) != 'nan']", "_____no_output_____" ], [ "producers.sort()\nproducers", "_____no_output_____" ], [ "input_file = amazon_walmart_all_path\noutput_file = 'amazon_walmart_output3.csv'\nsettings_file = 'amazon_walmart_learned_settings3'\ntraining_file = 'amazon_walmart_training3.json'", "_____no_output_____" ], [ "float('1.25')", "_____no_output_____" ], [ "def preProcess(key, column):\n \n try : # python 2/3 string differences\n column = column.decode('utf8')\n except AttributeError:\n pass\n column = unidecode(column)\n column = re.sub(' +', ' ', column)\n column = re.sub('\\n', ' ', column)\n column = column.strip().strip('\"').strip(\"'\").lower().strip()\n column = column.lower()\n if not column:\n return None\n \n if key == 'price':\n column = float(column) \n return column\n\ndef readData(filename):\n \n data_d = {}\n with open(filename) as f:\n reader = csv.DictReader(f)\n for row in reader:\n clean_row = [(k, preProcess(k, v)) for (k, v) in row.items()]\n row_id = int(row['Id'])\n data_d[row_id] = dict(clean_row)\n\n return data_d ", "_____no_output_____" ], [ "print('importing data ...')\ndata_d = readData(input_file)", "importing data ...\n" ], [ "fields = [\n {'field' : 'name', 'type': 'Name'},\n {'field' : 'name', 'type': 'String'},\n {'field' : 'description', \n 'type': 'Text',\n 'corpus': description_corpus,\n 'has_missing': True\n },\n {'field' : 'category', \n 'type': 'FuzzyCategorical',\n 'categories': categories,\n 'corpus': category_corpus,\n 'has missing' : True\n }, \n {'field' : 'producer', \n 'type': 'FuzzyCategorical',\n 'categories': producers,\n 'corpus': producer_corpus,\n 'has_missing': True\n },\n {'field' : 'price', \n 'type': 'Price',\n 'has_missing': True\n },\n]", "_____no_output_____" ], [ "deduper = dedupe.Dedupe(fields)", "_____no_output_____" ], [ "# took about 20 min with blocked proportion 0.8\ndeduper.prepare_training(data_d)", "INFO:dedupe.canopy_index:Removing stop word with\nINFO:dedupe.canopy_index:Removing stop word 7\nINFO:dedupe.canopy_index:Removing stop word is\nINFO:dedupe.canopy_index:Removing stop word than\nINFO:dedupe.canopy_index:Removing stop word get\nINFO:dedupe.canopy_index:Removing stop word quickly\nINFO:dedupe.canopy_index:Removing stop word that\nINFO:dedupe.canopy_index:Removing stop word your\nINFO:dedupe.canopy_index:Removing stop word into\nINFO:dedupe.canopy_index:Removing stop word and\nINFO:dedupe.canopy_index:Removing stop word share\nINFO:dedupe.canopy_index:Removing stop word works\nINFO:dedupe.canopy_index:Removing stop word you\nINFO:dedupe.canopy_index:Removing stop word protect\nINFO:dedupe.canopy_index:Removing stop word of\nINFO:dedupe.canopy_index:Removing stop word access\nINFO:dedupe.canopy_index:Removing stop word are\nINFO:dedupe.canopy_index:Removing stop word a\nINFO:dedupe.canopy_index:Removing stop word easy\nINFO:dedupe.canopy_index:Removing stop word to\nINFO:dedupe.canopy_index:Removing stop word off\nINFO:dedupe.canopy_index:Removing stop word wide\nINFO:dedupe.canopy_index:Removing stop word or\nINFO:dedupe.canopy_index:Removing stop word dvd\nINFO:dedupe.canopy_index:Removing stop word play\nINFO:dedupe.canopy_index:Removing stop word other\nINFO:dedupe.canopy_index:Removing stop word media\nINFO:dedupe.canopy_index:Removing stop word automatically\nINFO:dedupe.canopy_index:Removing stop word based\nINFO:dedupe.canopy_index:Removing stop word such\nINFO:dedupe.canopy_index:Removing stop word make\nINFO:dedupe.canopy_index:Removing stop word the\nINFO:dedupe.canopy_index:Removing stop word for\nINFO:dedupe.canopy_index:Removing stop word go\nINFO:dedupe.canopy_index:Removing stop word camera\nINFO:dedupe.canopy_index:Removing stop word in\nINFO:dedupe.canopy_index:Removing stop word create\nINFO:dedupe.canopy_index:Removing stop word just\nINFO:dedupe.canopy_index:Removing stop word using\nINFO:dedupe.canopy_index:Removing stop word all\nINFO:dedupe.canopy_index:Removing stop word use\nINFO:dedupe.canopy_index:Removing stop word video\nINFO:dedupe.canopy_index:Removing stop word them\nINFO:dedupe.canopy_index:Removing stop word audio\nINFO:dedupe.canopy_index:Removing stop word view\nINFO:dedupe.canopy_index:Removing stop word only\nINFO:dedupe.canopy_index:Removing stop word s\nINFO:dedupe.canopy_index:Removing stop word help\nINFO:dedupe.canopy_index:Removing stop word hard\nINFO:dedupe.canopy_index:Removing stop word up\nINFO:dedupe.canopy_index:Removing stop word plus\nINFO:dedupe.canopy_index:Removing stop word any\nINFO:dedupe.canopy_index:Removing stop word more\nINFO:dedupe.canopy_index:Removing stop word most\nINFO:dedupe.canopy_index:Removing stop word music\nINFO:dedupe.canopy_index:Removing stop word if\nINFO:dedupe.canopy_index:Removing stop word technology\nINFO:dedupe.canopy_index:Removing stop word high\nINFO:dedupe.canopy_index:Removing stop word screen\nINFO:dedupe.canopy_index:Removing stop word fast\nINFO:dedupe.canopy_index:Removing stop word simply\nINFO:dedupe.canopy_index:Removing stop word convenient\nINFO:dedupe.canopy_index:Removing stop word large\nINFO:dedupe.canopy_index:Removing stop word easily\nINFO:dedupe.canopy_index:Removing stop word right\nINFO:dedupe.canopy_index:Removing stop word type\nINFO:dedupe.canopy_index:Removing stop word style\nINFO:dedupe.canopy_index:Removing stop word look\nINFO:dedupe.canopy_index:Removing stop word what\nINFO:dedupe.canopy_index:Removing stop word files\nINFO:dedupe.canopy_index:Removing stop word designed\nINFO:dedupe.canopy_index:Removing stop word including\nINFO:dedupe.canopy_index:Removing stop word digital\nINFO:dedupe.canopy_index:Removing stop word multi\nINFO:dedupe.canopy_index:Removing stop word 30\nINFO:dedupe.canopy_index:Removing stop word when\nINFO:dedupe.canopy_index:Removing stop word 5\nINFO:dedupe.canopy_index:Removing stop word supports\nINFO:dedupe.canopy_index:Removing stop word compatible\nINFO:dedupe.canopy_index:Removing stop word small\nINFO:dedupe.canopy_index:Removing stop word x\nINFO:dedupe.canopy_index:Removing stop word viewing\nINFO:dedupe.canopy_index:Removing stop word ideal\nINFO:dedupe.canopy_index:Removing stop word this\nINFO:dedupe.canopy_index:Removing stop word great\nINFO:dedupe.canopy_index:Removing stop word design\nINFO:dedupe.canopy_index:Removing stop word space\nINFO:dedupe.canopy_index:Removing stop word users\nINFO:dedupe.canopy_index:Removing stop word features\nINFO:dedupe.canopy_index:Removing stop word usb\nINFO:dedupe.canopy_index:Removing stop word plug\nINFO:dedupe.canopy_index:Removing stop word comes\nINFO:dedupe.canopy_index:Removing stop word pc\nINFO:dedupe.canopy_index:Removing stop word 10\nINFO:dedupe.canopy_index:Removing stop word inch\nINFO:dedupe.canopy_index:Removing stop word t\nINFO:dedupe.canopy_index:Removing stop word case\nINFO:dedupe.canopy_index:Removing stop word 100\nINFO:dedupe.canopy_index:Removing stop word home\nINFO:dedupe.canopy_index:Removing stop word 2\nINFO:dedupe.canopy_index:Removing stop word network\nINFO:dedupe.canopy_index:Removing stop word its\nINFO:dedupe.canopy_index:Removing stop word photo\nINFO:dedupe.canopy_index:Removing stop word front\nINFO:dedupe.canopy_index:Removing stop word larger\nINFO:dedupe.canopy_index:Removing stop word allows\nINFO:dedupe.canopy_index:Removing stop word 4\nINFO:dedupe.canopy_index:Removing stop word photos\nINFO:dedupe.canopy_index:Removing stop word hours\nINFO:dedupe.canopy_index:Removing stop word work\nINFO:dedupe.canopy_index:Removing stop word over\nINFO:dedupe.canopy_index:Removing stop word products\nINFO:dedupe.canopy_index:Removing stop word applications\nINFO:dedupe.canopy_index:Removing stop word wireless\nINFO:dedupe.canopy_index:Removing stop word box\nINFO:dedupe.canopy_index:Removing stop word without\nINFO:dedupe.canopy_index:Removing stop word top\nINFO:dedupe.canopy_index:Removing stop word includes\nINFO:dedupe.canopy_index:Removing stop word interface\nINFO:dedupe.canopy_index:Removing stop word compact\nINFO:dedupe.canopy_index:Removing stop word desktop\nINFO:dedupe.canopy_index:Removing stop word also\nINFO:dedupe.canopy_index:Removing stop word support\nINFO:dedupe.canopy_index:Removing stop word they\nINFO:dedupe.canopy_index:Removing stop word offers\nINFO:dedupe.canopy_index:Removing stop word mac\nINFO:dedupe.canopy_index:Removing stop word key\nINFO:dedupe.canopy_index:Removing stop word data\nINFO:dedupe.canopy_index:Removing stop word built\nINFO:dedupe.canopy_index:Removing stop word power\nINFO:dedupe.canopy_index:Removing stop word not\nINFO:dedupe.canopy_index:Removing stop word player\nINFO:dedupe.canopy_index:Removing stop word 3\nINFO:dedupe.canopy_index:Removing stop word set\nINFO:dedupe.canopy_index:Removing stop word protection\nINFO:dedupe.canopy_index:Removing stop word back\nINFO:dedupe.canopy_index:Removing stop word file\nINFO:dedupe.canopy_index:Removing stop word year\nINFO:dedupe.canopy_index:Removing stop word warranty\nINFO:dedupe.canopy_index:Removing stop word about\nINFO:dedupe.canopy_index:Removing stop word re\nINFO:dedupe.canopy_index:Removing stop word fit\nINFO:dedupe.canopy_index:Removing stop word has\nINFO:dedupe.canopy_index:Removing stop word per\nINFO:dedupe.canopy_index:Removing stop word like\nINFO:dedupe.canopy_index:Removing stop word have\nINFO:dedupe.canopy_index:Removing stop word full\nINFO:dedupe.canopy_index:Removing stop word standard\nINFO:dedupe.canopy_index:Removing stop word maximum\nINFO:dedupe.canopy_index:Removing stop word 0\nINFO:dedupe.canopy_index:Removing stop word which\nINFO:dedupe.canopy_index:Removing stop word each\nINFO:dedupe.canopy_index:Removing stop word while\nINFO:dedupe.canopy_index:Removing stop word through\nINFO:dedupe.canopy_index:Removing stop word life\nINFO:dedupe.canopy_index:Removing stop word adapter\nINFO:dedupe.canopy_index:Removing stop word installation\nINFO:dedupe.canopy_index:Removing stop word cd\nINFO:dedupe.canopy_index:Removing stop word be\nINFO:dedupe.canopy_index:Removing stop word capacity\nINFO:dedupe.canopy_index:Removing stop word system\nINFO:dedupe.canopy_index:Removing stop word extra\nINFO:dedupe.canopy_index:Removing stop word dual\nINFO:dedupe.canopy_index:Removing stop word systems\nINFO:dedupe.canopy_index:Removing stop word keep\n" ], [ "dedupe.consoleLabel(deduper)", "name : durable bridge\ncategory : audio video accessories\nproducer : durable\nprice : 203.86\n\nname : durable bridge\ncategory : audio video accessories\nproducer : durable\nprice : 203.86\n\n0/10 positive, 0/10 negative\nDo these records refer to the same thing?\n(y)es / (n)o / (u)nsure / (f)inished\n" ], [ "data_d[1]", "_____no_output_____" ], [ "deduper.train()", "INFO:rlr.crossvalidation:using cross validation to find optimum alpha...\n/home/ubuntu/anaconda3/lib/python3.7/site-packages/rlr/crossvalidation.py:122: RuntimeWarning: invalid value encountered in double_scalars\n * (true_distinct + false_distinct)))\nINFO:rlr.crossvalidation:optimum alpha: 0.000100, score 0.24869807198976637\nINFO:dedupe.training:Final predicate set:\nINFO:dedupe.training:(SimplePredicate: (oneGramFingerprint, name), SimplePredicate: (sortedAcronym, name))\nINFO:dedupe.training:(SimplePredicate: (roundTo1, price), TfidfTextCanopyPredicate: (0.8, name))\n" ], [ "threshold = deduper.threshold(data_d, recall_weight=1)\nthreshold", "INFO:dedupe.canopy_index:Removing stop word new\nINFO:dedupe.canopy_index:Removing stop word 4\nINFO:dedupe.canopy_index:Removing stop word black\nINFO:dedupe.canopy_index:Removing stop word digital\nINFO:dedupe.canopy_index:Removing stop word with\nINFO:dedupe.canopy_index:Removing stop word and\nINFO:dedupe.canopy_index:Removing stop word white\nINFO:dedupe.canopy_index:Removing stop word case\nINFO:dedupe.canopy_index:Removing stop word x\nINFO:dedupe.canopy_index:Removing stop word inch\nINFO:dedupe.canopy_index:Removing stop word 1\nINFO:dedupe.canopy_index:Removing stop word for\nINFO:dedupe.canopy_index:Removing stop word gb\nINFO:dedupe.canopy_index:Removing stop word 2\nINFO:dedupe.canopy_index:Removing stop word usb\nINFO:dedupe.canopy_index:Removing stop word 8\nINFO:dedupe.canopy_index:Removing stop word 3\nINFO:dedupe.blocking:10000, 2.3534082 seconds\nINFO:dedupe.blocking:20000, 4.5438192 seconds\nINFO:dedupe.api:Maximum expected recall and precision\nINFO:dedupe.api:recall: 0.891\nINFO:dedupe.api:precision: 0.688\nINFO:dedupe.api:With threshold: 0.424\n" ], [ "print('clustering...')\nclustered_dupes = deduper.match(data_d, threshold)\n\nprint('# duplicate sets', len(clustered_dupes))", "clustering...\n" ], [ "for key, values in data_d.items():\n values['price'] = str(values['price']) ", "_____no_output_____" ], [ "cluster_membership = {}\ncluster_id = 0\nfor (cluster_id, cluster) in enumerate(clustered_dupes):\n id_set, scores = cluster\n cluster_d = [data_d[c] for c in id_set]\n \n \n canonical_rep = dedupe.canonicalize(cluster_d)\n for record_id, score in zip(id_set, scores):\n cluster_membership[record_id] = {\n \"cluster id\" : cluster_id,\n \"canonical representation\" : canonical_rep,\n \"confidence\": score\n }\n\nsingleton_id = cluster_id + 1\n\nwith open(output_file, 'w') as f_output, open(input_file) as f_input:\n writer = csv.writer(f_output)\n reader = csv.reader(f_input)\n\n heading_row = next(reader)\n heading_row.insert(0, 'confidence_score')\n heading_row.insert(0, 'Cluster ID')\n canonical_keys = canonical_rep.keys()\n for key in canonical_keys:\n heading_row.append('canonical_' + key)\n\n writer.writerow(heading_row)\n\n for row in reader:\n row_id = int(row[0])\n if row_id in cluster_membership:\n cluster_id = cluster_membership[row_id][\"cluster id\"]\n canonical_rep = cluster_membership[row_id][\"canonical representation\"]\n row.insert(0, cluster_membership[row_id]['confidence'])\n row.insert(0, cluster_id)\n for key in canonical_keys:\n row.append(canonical_rep[key].encode('utf8'))\n else:\n row.insert(0, None)\n row.insert(0, singleton_id)\n singleton_id += 1\n for key in canonical_keys:\n row.append(None)\n writer.writerow(row)\n ", "_____no_output_____" ], [ "fields_of_interest = ['Cluster ID', 'confidence_score', 'Id', 'name', 'producer', 'description', 'price']", "_____no_output_____" ], [ "amazon_walmart_output = pd.read_csv('amazon_walmart_output2.csv', sep=',', quotechar='\"')[fields_of_interest]", "_____no_output_____" ], [ "amazon_walmart_output[amazon_walmart_output['confidence_score'] == None]", "_____no_output_____" ], [ "amazon_walmart_output = amazon_walmart_output[fields_of_interest]", "_____no_output_____" ], [ "amazon_walmart_output[amazon_walmart_output['confidence_score'] > 0.9].sort_values('Cluster ID')", "_____no_output_____" ] ] ]

[ "code", "markdown", "code" ]

[ [ "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "%matplotlib inline\nimport io\nimport os\nimport matplotlib.pyplot as plt\nimport matplotlib.patches as patch\nimport pandas as pd\nimport numpy as np\nimport seaborn as sns\nsns.set_style('darkgrid')\nsns.set_context('notebook')", "c:\\users\\hjl161\\appdata\\local\\programs\\python\\python37\\lib\\site-packages\\statsmodels\\tools\\_testing.py:19: FutureWarning: pandas.util.testing is deprecated. Use the functions in the public API at pandas.testing instead.\n import pandas.util.testing as tm\n" ], [ "datafile_AMT = '../data/MTurk_anonymous.xlsx'\n#datafile_DTU1 = '../data/DTU1_anonymous.xlsx'\n#datafile_DTU2 = '../data/DTU2_anonymous.xlsx'", "_____no_output_____" ], [ "df_MTurk = pd.DataFrame(pd.read_excel(datafile_AMT))\ndf_MTurk.drop(df_MTurk.columns[df_MTurk.columns.str.contains('unnamed',case = False)],axis = 1, inplace = True)\n#df_DTU1 = pd.DataFrame(pd.read_excel(datafile_DTU1))\n#df_DTU1.drop(df_DTU1.columns[df_DTU1.columns.str.contains('unnamed',case = False)],axis = 1, inplace = True)\n#df_DTU2 = pd.DataFrame(pd.read_excel(datafile_DTU2))\n#df_DTU2.drop(df_DTU2.columns[df_DTU2.columns.str.contains('unnamed',case = False)],axis = 1, inplace = True)", "_____no_output_____" ], [ "df_MTurk.head()", "_____no_output_____" ] ], [ [ "In the following, we want to partition responses according to how many times a player has miscoordinated before. For a start, we concatenate all three experiments into one dataframe:", "_____no_output_____" ] ], [ [ "#df = pd.concat([\n# df_MTurk.join(pd.Series(['MTurk']*len(df_MTurk), name='experiment')), \n# df_DTU1.join(pd.Series(['DTU1']*len(df_DTU1), name='experiment')),\n# df_DTU2.join(pd.Series(['DTU2']*len(df_DTU2), name='experiment'))],\n# ignore_index=True)", "_____no_output_____" ], [ "#df.head(), len(df)", "_____no_output_____" ] ], [ [ "Now we group the data into response pairs", "_____no_output_____" ] ], [ [ "df = df_MTurk.groupby(['session', 'group', 'round'], as_index=False)[['code', 'id_in_group', 'arrival', 'choice', 'certainty']].agg(lambda x: tuple(x))\ndf.head()", "_____no_output_____" ], [ "#df0 = pd.DataFrame(columns=df.columns)\n#df1_ = pd.DataFrame(columns=df.columns)\n#sessions = df.session.unique()\n#for session in sessions:\n# df_session = df[df.session == session]\n# groups = df_session.group.unique()\n# for group in groups:\n# df_group = df_session[df_session.group == group]\n# miss = 0\n# for idx, row in df.iterrows():\n# if sum(row['choice']) != 1:\n# row['miss'] = miss\n# df0 = df0.append(row, ignore_index=True)\n# else:\n# miss += 1\n# df1_ = df1_.append(row, ignore_index=True)\n#df.head()", "_____no_output_____" ], [ "# initialize new dataframes that will hold data with a condition. df0 will contain all choices \n# having had zero previous miscoordinations, while df1_ will contain all choices having had one\n# or more previous miscoordinations\ndf0 = pd.DataFrame(columns=df.columns)\ndf1_ = pd.DataFrame(columns=df.columns)", "_____no_output_____" ], [ "# partition the dataframe into two bins - the first having 0 and 1, and the other 2 or more\n# miscoordinations:\nsessions = df.session.unique()\nfor session in sessions:\n df_session = df[df.session == session]\n groups = df_session.group.unique()\n for group in groups:\n df_group = df_session[df_session.group == group]\n miss = 0\n for idx, row in df_group.iterrows():\n if sum(row['choice']) != 1:\n if miss == 0:\n df0 = df0.append(row, ignore_index=True)\n else:\n df1_ = df1_.append(row, ignore_index=True)\n else:\n if miss == 0:\n df0 = df0.append(row, ignore_index=True)\n else:\n df1_ = df1_.append(row, ignore_index=True)\n miss += 1", "_____no_output_____" ] ], [ [ "Now a bit of magic: we need to separate the tuples again and create rows for each person: (see ref at https://stackoverflow.com/questions/53218931/how-to-unnest-explode-a-column-in-a-pandas-dataframe", "_____no_output_____" ] ], [ [ "def unnesting(df, explode):\n idx = df.index.repeat(df[explode[0]].str.len())\n df1 = pd.concat([\n pd.DataFrame({x: np.concatenate(df[x].values)}) for x in explode], axis=1)\n df1.index = idx\n\n return df1.join(df.drop(explode, 1), how='left')\n\n\ndfnew0 = unnesting(df0,['code', 'id_in_group', 'arrival', 'choice', 'certainty'])\ndfnew1_ = unnesting(df1_,['code', 'id_in_group', 'arrival', 'choice', 'certainty'])", "_____no_output_____" ], [ "dfnew0['arrival'].replace({9.0: 8.6, 9.1: 8.7}, inplace=True)\ndfnew1_['arrival'].replace({9.0: 8.6, 9.1: 8.7}, inplace=True)\nlen(dfnew0), len(dfnew1_)", "_____no_output_____" ], [ "dfnew0.head()", "_____no_output_____" ], [ "sns.regplot(x='arrival', y='choice', data=dfnew0, ci=95, logistic=True)\nsns.despine()", "_____no_output_____" ], [ "sns.regplot(x='arrival', y='choice', data=dfnew1_, ci=95, logistic=True)\nsns.despine()", "_____no_output_____" ], [ "dfall = pd.concat([dfnew0.join(pd.Series(['zero']*len(dfnew0), name='miss')), \n dfnew1_.join(pd.Series(['one_or_more']*len(dfnew1_), name='miss'))],\n ignore_index=True)", "_____no_output_____" ], [ "dfall.head()", "_____no_output_____" ], [ "pal = dict(zero=\"blue\", one_or_more=\"orange\")\ng = sns.lmplot(x=\"arrival\", y=\"choice\", hue=\"miss\", data=dfall, palette=pal, \n logistic=True, ci=95, n_boot=10000, x_estimator=np.mean, x_ci=\"ci\",\n y_jitter=.2, legend=False, height=3, aspect=1.5)\n#g = sns.FacetGrid(dfall, hue='miscoordinations', palette=pal, height=5);\n#g.map(plt.scatter, \"arrival\", \"choice\", marker='o', s=50, alpha=.7, linewidth=.5, color='white', edgecolor=\"white\")\n#g.map(sns.regplot, \"arrival\", \"choice\", scatter_kws={\"color\": \"grey\"}, ci=95, n_boot=100, logistic=True, height=3, aspect=1.5)\ng.set(xlim=(7.98, 8.72))\ng.set(ylabel='frequency of canteen choice')\nmy_ticks = [\"8:00\", \"8:10\", \"8:20\", \"8:30\", \"8:40\", \"8:50\", \"9:00\", \"9:10\"]\ng.set(xticks=[8,8.1,8.2,8.3,8.4,8.5,8.6,8.7], yticks=[0,.1,.2,.3,.4,.5,.6,.7,.8,.9,1])\ng.set(xticklabels = my_ticks) #, yticklabels = [\"office\",\"\",\"\",\"\",\"\",\".5\",\"\",\"\",\"\",\"\",\"canteen\"])\n\n# make my own legend:\nname_to_color = {\n ' 0 (n=2762)': 'blue',\n '>0 (n=1498)': 'orange',\n}\npatches = [patch.Patch(color=v, label=k) for k,v in name_to_color.items()]\nplt.legend(title='# miscoordinations', handles=patches, loc='lower left')\nplt.title('MTurk', fontsize=16)\nplt.tight_layout()\n\nplt.rcParams[\"font.family\"] = \"sans-serif\"\nPLOTS_DIR = '../plots'\n\nif not os.path.exists(PLOTS_DIR):\n os.makedirs(PLOTS_DIR)\n\nplt.savefig(os.path.join(PLOTS_DIR, 'figS6_logit_2bins_MTurk.png'),\n bbox_inches='tight', transparent=True, dpi=300)\nplt.savefig(os.path.join(PLOTS_DIR, 'figS6_logit_2bins_MTurk.pdf'), transparent=True, dpi=300)\n\nsns.despine()", "_____no_output_____" ] ] ]

[ "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT-0" ]

[ "MIT-0" ]

[ "MIT-0" ]

[ [ [ "# COVID-MultiVent\n\nThe COVID MultiVent System was developed through a collaboration between Dr. Arun Agarwal, pulmonary critical care fellow at the University of Missouri, and a group of medical students at the University of Pennsylvania with backgrounds in various types of engineering. The system allows for monitoring of pressure and volume delivered to each patient and incorporates valve components that allow for peak inspiratory pressure (PIP), positive end expiratory pressure (PEEP), and inspiratory volume to be modulated independently for each patient. We have created open source assembly resources for this system to expedite its adoption in hospitals around the world with dire need of ventilation resources. The system is assembled with components that can be readily found in most hospitals. If you have questions regarding the system please contact us at **[email protected]** \n\nDr. Arun Agarwal\n\nAlexander Lee\n\nFeyisope Eweje\n\nStephen Landy\n\nDavid Gordon\n\nRyan Gallagher \n\nAlfredo Lucas", "_____no_output_____" ] ], [ [ "%%html\n<iframe src=\"https://docs.google.com/presentation/d/e/2PACX-1vT2kAsF54vHJoIti5iQkk8zUTkcQuRBFnbtgAhCoHwxrThIzZ14mCpdgNkWrqS8bRf5xFB2ZoeXlfUk/embed?start=false&loop=false&delayms=3000\" frameborder=\"0\" width=\"480\" height=\"299\" allowfullscreen=\"true\" mozallowfullscreen=\"true\" webkitallowfullscreen=\"true\"></iframe>", "_____no_output_____" ] ], [ [ "### Original Video and Facebook Group", "_____no_output_____" ], [ "Please see below Dr. Agrawal's original video showing the proposed setup. A link to the MultiVent facebook group can be found [here](https://facebook.com/COVIDMULTIVENT/).", "_____no_output_____" ] ], [ [ "%%html\n\n<iframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/odnhnnlBlpM\" frameborder=\"0\" allow=\"accelerometer; autoplay; encrypted-media; gyroscope; picture-in-picture\" allowfullscreen></iframe>", "_____no_output_____" ] ], [ [ "# Assembly Guide and Component List\n\nFor detailed instructions on how to build the MultiVent setup please click [here](https://docs.google.com/document/d/11Z7MpeGSC6m3ipsMVLe-Ofkw4HeqCN6x1Zhqy8FMRpc/edit?usp=sharing)\n\n\nBelow see the proposed parts list with the number of each part required for a single or a 2-patient setup. A list of vendors associated with each part is also included, however, note that almost all of these components will be readily available in many hospitals that already have ventilator setups. \nClick [here](https://docs.google.com/spreadsheets/d/1XikdFKNdgZAywoPw4oIWqJ9e0-a059ucXh8DSOGl4GA/edit?usp=sharing) for a Google Sheets version of this list.", "_____no_output_____" ] ], [ [ "import pandas as pd\n\ndf = pd.read_excel('Website Ventilator splitting components.xlsx')\ndf", "_____no_output_____" ] ], [ [ "## Additional Resources", "_____no_output_____" ], [ "[US Public Health Service guidelines for ventilator splitting](https://www.hhs.gov/sites/default/files/optimizing-ventilator-use-during-covid19-pandemic.pdf)\n\n[Ventilation Sharing Protocol developed by the Columbia University College of Physicians & Surgeons/New York-Presbyterian Hospital (3/24/20)](https://www.gnyha.org/wp-content/uploads/2020/03/Ventilator-Sharing-Protocol-Dual-Patient-Ventilation-with-a-Single-Mechanical-Ventilator-for-Use-during-Critical-Ventilator-Shortages.pdf)\n\n[PulmCrit - Splitting ventilators to provide titrated support to a large group of patients](https://emcrit.org/pulmcrit/split-ventilators/)\n\n[Medium article explain vent splitting methods - A better way of connecting multiple patients to a single ventilator](https://medium.com/@pinsonhannah/a-better-way-of-connecting-multiple-patients-to-a-single-ventilator-fa9cf42679c6)\n", "_____no_output_____" ], [ "# ***DISCLAIMER - This system is currently not endorsed by the University of Missouri nor the University of Pennsylvania. Multi-patient ventilation should be attempted with the greatest of caution as an option of last resort with appropriate clinical guidance. The COVID MultiVent team does not assume any liability for adverse events that may occur that could in any way be tied to use of the system.***\n\n", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "import glob\nimport itertools\nfrom ipywidgets import widgets, Layout\nimport numpy as np\nimport os\nimport pandas as pd\nimport plotly.io as pio\nimport plotly.graph_objects as go\n\nfrom apex_performance_plotter.apex_performance_plotter.load_logfiles import load_logfiles\n\npio.templates.default = \"plotly_white\"\nfrom IPython.core.interactiveshell import InteractiveShell", "_____no_output_____" ], [ "# Define the folders where to look for experiment outputs\nos.chdir('../../../../experiment')\nlogfiles = glob.glob('{}*'.format('log'))\nselected_logfiles = widgets.SelectMultiple(\n options=logfiles,\n description='Experiments',\n disabled=False,\n layout=Layout(width='100%')\n)\ndisplay(selected_logfiles)\n\n# Select the experiments to plot", "_____no_output_____" ], [ "# Display selected experiment properties\nInteractiveShell.ast_node_interactivity = \"all\"\n\nheaders, dataframes = load_logfiles(selected_logfiles)\n\nfor idx, header in enumerate(headers):\n display(header)\n \nInteractiveShell.ast_node_interactivity = \"last\"", "_____no_output_____" ], [ "colors = ['#4363d8','#800000','#f58231','#e6beff']\n\n# Plot latencies\nfigure_latencies = go.FigureWidget()\nfigure_latencies.layout.xaxis.title = 'Time (s)'\nfigure_latencies.layout.yaxis.title = 'Latencies (ms)'\n\nfor idx, experiment in enumerate(dataframes):\n\n figure_latencies.add_scatter(x=experiment['T_experiment'],\n y=experiment['latency_max (ms)'],\n mode='markers', marker_color=colors[idx],\n marker_symbol='x',\n name= 'latency_max',\n text=headers[idx]['Logfile name']);\n figure_latencies.add_scatter(x=experiment['T_experiment'],\n y=experiment['latency_mean (ms)'],\n mode='markers', marker_color=colors[idx],\n marker_symbol='triangle-up',\n name='latency_mean',\n text=headers[idx]['Logfile name']);\n figure_latencies.add_scatter(x=experiment['T_experiment'],\n y=experiment['latency_min (ms)'],\n mode='markers', marker_color=colors[idx],\n name='latency_min',\n text=headers[idx]['Logfile name'])\n\nfigure_latencies", "_____no_output_____" ], [ "# Plot CPU usage\nfigure_cpu_usage = go.FigureWidget()\nfigure_cpu_usage.layout.xaxis.title = 'Time (s)'\nfigure_cpu_usage.layout.yaxis.title = 'CPU usage (%)'\n\nfor idx, experiment in enumerate(dataframes):\n\n figure_cpu_usage.add_scatter(x=experiment['T_experiment'],\n y=experiment['cpu_usage (%)'],\n mode='markers', marker_color=colors[idx],\n marker_symbol='x',\n name= 'cpu_usage',\n text=headers[idx]['Logfile name']);\n\nfigure_cpu_usage", "_____no_output_____" ], [ "# Plot memory consumption\nfigure_memory_usage = go.FigureWidget()\nfigure_memory_usage.layout.xaxis.title = 'Time (s)'\nfigure_memory_usage.layout.yaxis.title = 'Memory consumption (MB)'\n\nfor idx, experiment in enumerate(dataframes):\n\n figure_memory_usage.add_scatter(x=experiment['T_experiment'],\n y=experiment['ru_maxrss']/1024,\n mode='markers', marker_color=colors[idx],\n marker_symbol='x',\n name= 'ru_maxrss',\n text=headers[idx]['Logfile name']);\n\nfigure_memory_usage", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Homework 10: `SQL`\n\n## Due Date: Thursday, November 16th at 11:59 PM\n\nYou will create a database of the NASA polynomial coefficients for each specie.\n\n**Please turn in your database with your `Jupyter` notebook!**", "_____no_output_____" ], [ "# Question 1: Convert XML to a SQL database", "_____no_output_____" ], [ "Create two tables named `LOW` and `HIGH`, each corresponding to data given for the low and high temperature range.\nEach should have the following column names:\n\n- `SPECIES_NAME`\n- `TLOW`\n- `THIGH`\n- `COEFF_1`\n- `COEFF_2`\n- `COEFF_3`\n- `COEFF_4`\n- `COEFF_5`\n- `COEFF_6`\n- `COEFF_7`\n\nPopulate the tables using the XML file you created in last assignment. If you did not complete the last assignment, you may also use the `example_thermo.xml` file.\n\n`TLOW` should refer to the temperature at the low range and `THIGH` should refer to the temperature at the high range. For example, in the `LOW` table, $H$ would have `TLOW` at $200$ and `THIGH` at $1000$ and in the `HIGH` table, $H$ would have `TLOW` at $1000$ and `THIGH` at $3500$.\n\nFor both tables, `COEFF_1` through `COEFF_7` should be populated with the corresponding coefficients for the low temperature data and high temperature data.", "_____no_output_____" ] ], [ [ "import xml.etree.ElementTree as ET\nimport sqlite3\nimport pandas as pd\nfrom IPython.core.display import display\n\npd.set_option('display.width', 500)\npd.set_option('display.max_columns', 100)\npd.set_option('display.notebook_repr_html', True)\n\n# Create and connect to database\ndb = sqlite3.connect('NASA_coef.sqlite')\ncursor = db.cursor()\ncursor.execute(\"DROP TABLE IF EXISTS LOW\")\ncursor.execute(\"DROP TABLE IF EXISTS HIGH\")\ncursor.execute(\"PRAGMA foreign_keys=1\")\n\n# Create the table for low temperature range\ncursor.execute('''CREATE TABLE LOW (\n id INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL, \n SPECIES_NAME TEXT,\n TLOW REAL, \n THIGH REAL, \n COEFF_1 REAL, \n COEFF_2 REAL,\n COEFF_3 REAL,\n COEFF_4 REAL,\n COEFF_5 REAL,\n COEFF_6 REAL,\n COEFF_7 REAL)''')\n\ndb.commit()\n\n# Create the table for high temperature range\ncursor.execute('''CREATE TABLE HIGH (\n id INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL, \n SPECIES_NAME TEXT,\n TLOW REAL, \n THIGH REAL, \n COEFF_1 REAL, \n COEFF_2 REAL,\n COEFF_3 REAL,\n COEFF_4 REAL,\n COEFF_5 REAL,\n COEFF_6 REAL,\n COEFF_7 REAL)''')\ndb.commit()\n\n# The given helper function (from L18) to visualize tables\ndef viz_tables(cols, query):\n q = cursor.execute(query).fetchall()\n framelist = []\n for i, col_name in enumerate(cols):\n framelist.append((col_name, [col[i] for col in q]))\n return pd.DataFrame.from_items(framelist)\n\n", "_____no_output_____" ], [ "tree = ET.parse('thermo.xml')\n\nspecies_data = tree.find('speciesData')\nspecies_list = species_data.findall('species') # a list of all species\n\nfor species in species_list:\n name = species.get('name')\n NASA_list = species.find('thermo').findall('NASA')\n for NASA in NASA_list:\n T_min = float(NASA.get('Tmin'))\n T_max = float(NASA.get('Tmax'))\n coefs = NASA.find('floatArray').text.split(',')\n vals_to_insert = (name, T_min, T_max, float(coefs[0].strip()), float(coefs[1].strip()), \n float(coefs[2].strip()), float(coefs[3].strip()), float(coefs[4].strip()), \n float(coefs[5].strip()), float(coefs[6].strip())) \n\n if T_max > 1000: # high range temperature\n cursor.execute('''INSERT INTO HIGH \n (SPECIES_NAME, TLOW, THIGH, COEFF_1, COEFF_2, COEFF_3, COEFF_4, COEFF_5, COEFF_6, COEFF_7)\n VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?)''', vals_to_insert)\n \n else: # low range temperature\n cursor.execute('''INSERT INTO LOW \n (SPECIES_NAME, TLOW, THIGH, COEFF_1, COEFF_2, COEFF_3, COEFF_4, COEFF_5, COEFF_6, COEFF_7)\n VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?)''', vals_to_insert)", "_____no_output_____" ], [ "LOW_cols = [col[1] for col in cursor.execute(\"PRAGMA table_info(LOW)\")]\ndisplay(viz_tables(LOW_cols, '''SELECT * FROM LOW'''))", "_____no_output_____" ], [ "HIGH_cols = [col[1] for col in cursor.execute(\"PRAGMA table_info(HIGH)\")]\ndisplay(viz_tables(HIGH_cols, '''SELECT * FROM HIGH'''))", "_____no_output_____" ] ], [ [ "# Question 2: `WHERE` Statements", "_____no_output_____" ], [ "1. Write a `Python` function `get_coeffs` that returns an array of 7 coefficients. \n \n The function should take in two parameters: 1.) `species_name` and 2.) `temp_range`, an indicator variable ('low' or 'high') to indicate whether the coefficients should come from the low or high temperature range. \n The function should use `SQL` commands and `WHERE` statements on the table you just created in Question 1 (rather than taking data from the XML directly).\n```python\ndef get_coeffs(species_name, temp_range):\n ''' Fill in here'''\n return coeffs\n```\n\n2. Write a python function `get_species` that returns all species that have a temperature range above or below a given value. The function should take in two parameters: 1.) `temp` and 2.) `temp_range`, an indicator variable ('low' or 'high').\n\n When temp_range is 'low', we are looking for species with a temperature range lower than the given temperature, and for a 'high' temp_range, we want species with a temperature range higher than the given temperature.\n\n This exercise may be useful if different species have different `LOW` and `HIGH` ranges.\n\n And as before, you should accomplish this through `SQL` queries and where statements.\n\n```python\ndef get_species(temp, temp_range):\n ''' Fill in here'''\n return coeffs\n```", "_____no_output_____" ] ], [ [ "def get_coeffs(species_name, temp_range):\n query = '''SELECT COEFF_1, COEFF_2, COEFF_3, COEFF_4, COEFF_5, COEFF_6, COEFF_7\n FROM {} \n WHERE SPECIES_NAME = \"{}\"'''.format(temp_range.upper(), species_name)\n coeffs = list(cursor.execute(query).fetchall()[0])\n return coeffs", "_____no_output_____" ], [ "get_coeffs('H', 'low')", "_____no_output_____" ], [ "def get_species(temp, temp_range):\n if temp_range == 'low': # temp_range == 'low'\n query = '''SELECT SPECIES_NAME FROM {} WHERE TLOW < {}'''.format(temp_range.upper(), temp)\n else: # temp_range == 'high'\n query = '''SELECT SPECIES_NAME FROM {} WHERE THIGH > {}'''.format(temp_range.upper(), temp)\n species = []\n for s in cursor.execute(query).fetchall():\n species.append(s[0])\n return species", "_____no_output_____" ], [ "get_species(500, 'low')", "_____no_output_____" ], [ "get_species(100, 'low')", "_____no_output_____" ], [ "get_species(3000, 'high')", "_____no_output_____" ], [ "get_species(3500, 'high')", "_____no_output_____" ] ], [ [ "# Question 3: `JOIN` STATEMENTS", "_____no_output_____" ], [ "Create a table named `ALL_TEMPS` that has the following columns:\n\n- `SPECIES_NAME`\n- `TEMP_LOW`\n- `TEMP_HIGH`\n\nThis table should be created by joining the tables `LOW` and `HIGH` on the value `SPECIES_NAME`.", "_____no_output_____" ] ], [ [ "# Create the table for ALL_TEMPS\ncursor.execute(\"DROP TABLE IF EXISTS ALL_TEMPS\")\n\ncursor.execute('''CREATE TABLE ALL_TEMPS (\n id INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL, \n SPECIES_NAME TEXT,\n TEMP_LOW REAL, \n TEMP_HIGH REAL)''')\ndb.commit()\n\n# Insert TEMP_LOW and TEMP_HIGH of all species to ALL_TEMPS\nquery = '''SELECT LOW.SPECIES_NAME, LOW.TLOW AS TEMP_LOW, HIGH.THIGH AS TEMP_HIGH\n FROM LOW\n INNER JOIN HIGH ON LOW.SPECIES_NAME = HIGH.SPECIES_NAME'''\n\nfor record in cursor.execute(query).fetchall():\n cursor.execute('''INSERT INTO ALL_TEMPS \n (SPECIES_NAME, TEMP_LOW, TEMP_HIGH)\n VALUES (?, ?, ?)''', record)", "_____no_output_____" ], [ "ALL_TEMPS_cols = [col[1] for col in cursor.execute(\"PRAGMA table_info(ALL_TEMPS)\")]\ndisplay(viz_tables(ALL_TEMPS_cols, '''SELECT * FROM ALL_TEMPS'''))", "_____no_output_____" ] ], [ [ "1. Write a `Python` function `get_range` that returns the range of temperatures for a given species_name.\n\nThe range should be computed within the `SQL` query (i.e. you should subtract within the `SELECT` statement in the `SQL` query).\n```python\ndef get_range(species_name):\n '''Fill in here'''\n return range\n```\n\nNote that `TEMP_LOW` is the lowest temperature in the `LOW` range and `TEMP_HIGH` is the highest temperature in the `HIGH` range.", "_____no_output_____" ] ], [ [ "def get_range(species_name):\n query = '''SELECT (TEMP_HIGH - TEMP_LOW) AS T_range FROM ALL_TEMPS WHERE SPECIES_NAME = \"{}\"'''.format(species_name)\n T_range = cursor.execute(query).fetchall()[0][0]\n return T_range", "_____no_output_____" ], [ "get_range('O')", "_____no_output_____" ], [ "# Close the Database\ndb.commit()\ndb.close()", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown", "markdown" ], [ "code", "code", "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code", "code", "code", "code", "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Results for BERT when applying syn tranformation to both premise and hypothesis to the MNLI dataset", "_____no_output_____" ] ], [ [ "import pandas as pd\nimport os\nimport numpy as np\nimport matplotlib.pyplot as plt\nimport seaborn as sns\nfrom tqdm import tqdm\nfrom IPython.display import display, HTML \nfrom lr.analysis.util import get_ts_from_results_folder \nfrom lr.analysis.util import get_rho_stats_from_result_list\nfrom lr.stats.h_testing import get_ks_stats_from_p_values_compared_to_uniform_dist", "_____no_output_____" ] ], [ [ "## Get Results", "_____no_output_____" ] ], [ [ "all_accs = []\nall_transformed_accs = []\nall_paired_t_p_values = []\nall_dev_plus_diff = []\nall_time = []\n\nm_name = \"bert_base\"\nfolder = \"mnli\"\ntest_repetitions = 5\nbatchs = range(1, test_repetitions + 1)\n\nfor i in tqdm(batchs):\n test_accuracy = get_ts_from_results_folder(path=\"results/{}/{}/syn_p_h/batch{}/\".format(folder,m_name, i),\n stat=\"test_accuracy\")\n \n transformed_test_accuracy = get_ts_from_results_folder(path=\"results/{}/{}/syn_p_h/batch{}/\".format(folder,m_name, i),\n stat=\"transformed_test_accuracy\")\n \n paired_t_p_value = get_ts_from_results_folder(path=\"results/{}/{}/syn_p_h/batch{}/\".format(folder, m_name, i),\n stat=\"paired_t_p_value\")\n \n diff = get_ts_from_results_folder(path=\"results/{}/{}/syn_p_h/batch{}/\".format(folder, m_name, i),\n stat=\"dev_plus_accuracy_difference\")\n \n t_time = get_ts_from_results_folder(path=\"results/{}/{}/syn_p_h/batch{}/\".format(folder, m_name,i),\n stat=\"test_time\")\n\n \n all_accs.append(test_accuracy)\n all_transformed_accs.append(transformed_test_accuracy)\n all_paired_t_p_values.append(paired_t_p_value)\n all_dev_plus_diff.append(diff)\n all_time.append(t_time)", "100%|██████████| 5/5 [00:00<00:00, 10.57it/s]\n" ], [ "total_time = pd.concat(all_time,1).sum().sum()\nn_params = 109484547\nprint(\"Time for all experiments = {:.1f} hours\".format(total_time))\nprint(\"Number of paramaters for BERT = {}\".format(n_params))", "Time for all experiments = 204.6 hours\nNumber of paramaters for BERT = 109484547\n" ] ], [ [ "## Accuracy", "_____no_output_____" ] ], [ [ "rhos, mean_acc, error_acc, _ = get_rho_stats_from_result_list(all_accs)\n\n_, mean_acc_t, error_acc_t, _ = get_rho_stats_from_result_list(all_transformed_accs)\n\nfig, ax = plt.subplots(figsize=(12,6))\nax.errorbar(rhos, mean_acc, yerr=error_acc, fmt='-o', label=\"original test data\");\nax.errorbar(rhos, mean_acc_t, yerr=error_acc_t, fmt='-o', label=\"transformed test data\");\nax.legend(loc=\"best\");\nax.set_xlabel(r\"$\\rho$\", fontsize=14);\nax.set_ylabel(\"accuracy\", fontsize=14);\nax.set_title(\"BERT accuracy\\n\\ndataset: MNLI\\ntransformation: synonym substitution\\ntest repetitions: {}\\n\".format(test_repetitions));\nfig.tight_layout()\nfig.savefig('figs/bert_base_acc_mnli_syn_p_h.png', bbox_inches=None, pad_inches=0.5)", "_____no_output_____" ] ], [ [ "## P-values", "_____no_output_____" ] ], [ [ "rhos, mean_p_values, error_p_values, min_p_values = get_rho_stats_from_result_list(all_paired_t_p_values)\n\nalpha = 0.05\nalpha_adj = alpha / test_repetitions\n\nrejected_ids = []\nremain_ids = []\n\nfor i,p in enumerate(min_p_values):\n if p < alpha_adj:\n rejected_ids.append(i)\n else:\n remain_ids.append(i)\n \nrhos_rejected = rhos[rejected_ids]\nrhos_remain = rhos[remain_ids]\ny_rejected = mean_p_values[rejected_ids]\ny_remain = mean_p_values[remain_ids]\nerror_rejected = error_p_values[rejected_ids]\nerror_remain = error_p_values[remain_ids]\n\ntitle_msg = \"BERT p-values\\n\\ndataset:\"\ntitle_msg += \"MNLI\\ntransformation: synonym substitution\\ntest repetitions: {}\\n\".format(test_repetitions)\ntitle_msg += \"significance level = {:.1%} \\n\".format(alpha)\ntitle_msg += \"adjusted significance level = {:.2%} \\n\".format(alpha_adj)\n\n\nfig, ax = plt.subplots(figsize=(12,6))\nax.errorbar(rhos_rejected, y_rejected, yerr=error_rejected, fmt='o', linewidth=0.50, label=\"at least one p-value is smaller than {:.2%}\".format(alpha_adj));\nax.errorbar(rhos_remain, y_remain, yerr=error_remain, fmt='o', linewidth=0.50, label=\"all p-values are greater than {:.2%}\".format(alpha_adj));\nax.legend(loc=\"best\");\nax.set_xlabel(r\"$\\rho$\", fontsize=14);\nax.set_ylabel(\"p-value\", fontsize=14);\nax.set_title(title_msg);\nfig.tight_layout()\nfig.tight_layout()\nfig.savefig('figs/bert_p_values_mnli_syn_p_h.png', bbox_inches=None, pad_inches=0.5)", "_____no_output_____" ] ], [ [ "## Accuracy difference", "_____no_output_____" ] ], [ [ "rhos, diff, _,_ = get_rho_stats_from_result_list(all_dev_plus_diff)\n_, test_acc, _,_ = get_rho_stats_from_result_list(all_accs)\n_, test_acc_t, _,_ = get_rho_stats_from_result_list(all_transformed_accs)\ntest_diff = np.abs(test_acc - test_acc_t)\n\nfig, ax = plt.subplots(figsize=(12,6))\nax.errorbar(rhos, diff, fmt='-o', label=\"validation\");\nax.errorbar(rhos, test_diff, fmt='-o', label=\"test\");\n\nax.legend(loc=\"best\");\nax.set_xlabel(r\"$\\rho$\", fontsize=14);\nax.set_ylabel(\"average accuracy difference\", fontsize=14);\nax.set_title(\"BERT accuracy difference\\n\\ndataset: MNLI\\ntransformation: synonym substitution\\ntest repetitions: {}\\n\".format(test_repetitions));\nfig.tight_layout()\nfig.savefig('figs/bert_acc_diff_mnli_syn_p_h.png', bbox_inches=None, pad_inches=0.5)", "_____no_output_____" ] ], [ [ "## Selecting the best $\\rho$", "_____no_output_____" ] ], [ [ "id_min = np.argmin(diff)\nmin_rho = rhos[id_min]\nmin_rho_test_acc = test_acc[id_min]\nmin_rho_transformed_test_acc = test_acc_t[id_min]\ntest_accuracy_loss_pct = np.round(((min_rho_test_acc - test_acc[0]) / test_acc[0]) * 100, 1)\n\nanalysis = {\"dataset\":\"mnli\",\n \"model\": \"BERT\",\n \"rho\":min_rho,\n \"test_accuracy_loss_pct\": test_accuracy_loss_pct,\n \"average_test_accuracy\": min_rho_test_acc,\n \"average_transformed_test_accuracy\": min_rho_transformed_test_acc,\n \"combined_accuracy\": np.mean([min_rho_test_acc,min_rho_transformed_test_acc])}\nanalysis = pd.DataFrame(analysis, index=[0])\nanalysis", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "[View in Colaboratory](https://colab.research.google.com/github/tompollard/buenosaires2018/blob/master/1_intro_to_python.ipynb)", "_____no_output_____" ], [ "# Programming in Python", "_____no_output_____" ], [ "In this part of the workshop, we will introduce some basic programming concepts in Python. We will then explore how these concepts allow us to carry out an anlysis that can be reproduced.", "_____no_output_____" ], [ "## Working with variables", "_____no_output_____" ], [ "You can get output from Python by typing math into a code cell. Try executing a sum below (for example: 3 + 5).", "_____no_output_____" ] ], [ [ "3+5", "_____no_output_____" ] ], [ [ "However, to do anything useful, we will need to assign values to `variables`. Assign a height in cm to a variable in the cell below.", "_____no_output_____" ] ], [ [ "height_cm = 180", "_____no_output_____" ] ], [ [ "Now the value has been assigned to our variable, we can print it in the console with `print`.", "_____no_output_____" ] ], [ [ "print('Height in cm is:', height_cm)", "_____no_output_____" ] ], [ [ "We can also do arithmetic with the variable. Convert the height in cm to metres, then print the new value as before (Warning! In Python 2, dividing an integer by an integer will return an integer.)", "_____no_output_____" ] ], [ [ "height_m = height_cm / 100\nprint('height in metres:',height_m)", "_____no_output_____" ] ], [ [ "We can check which variables are available in memory with the special command: `%whos`", "_____no_output_____" ] ], [ [ "%whos", "_____no_output_____" ] ], [ [ "We can see that each of our variables has a type (in this case `int` and `float`), describing the type of data held by the variable. We can use `type` to check the data type of a variable.", "_____no_output_____" ] ], [ [ "type(height_cm)", "_____no_output_____" ] ], [ [ "Another data type is a `list`, which can hold a series of items. For example, we might measure a patient's heart rate several times over a period. ", "_____no_output_____" ] ], [ [ "heartrate = [66,64,63,62,66,69,70,75,76]", "_____no_output_____" ], [ "print(heartrate)", "_____no_output_____" ], [ "type(heartrate)", "_____no_output_____" ] ], [ [ "## Repeating actions in loops", "_____no_output_____" ], [ "We can access individual items in a list using an index (note, in Python, indexing begins with 0!). For example, let's view the first `[0]` and second `[1]` heart rate measurements.", "_____no_output_____" ] ], [ [ "print(heartrate[0])", "_____no_output_____" ], [ "print(heartrate[1])", "_____no_output_____" ] ], [ [ "We can iterate through a list with the help of a `for` loop. Let's try looping over our list of heart rates, printing each item as we go.", "_____no_output_____" ] ], [ [ "for i in heartrate:\n print('the heart rate is:', i)", "_____no_output_____" ] ], [ [ "## Making choices", "_____no_output_____" ], [ "Sometimes we want to take different actions depending on a set of conditions. We can do this using an `if/else` statement. Let's write a statement to test if a mean arterial pressure (`meanpressure`) is high or low.", "_____no_output_____" ] ], [ [ "meanpressure = 70\n\nif meanpressure < 60:\n print('Low pressure')\nelif meanpressure > 100:\n print('High pressure')\nelse:\n print('Normal pressure')", "_____no_output_____" ] ], [ [ "## Writing our own functions", "_____no_output_____" ], [ "To help organise our code and to avoid replicating the same code again and again, we can create functions. \n\nLet's create a function to convert temperature in fahrenheit to celsius, using the following formula:\n\n`celsius = (fahrenheit - 32) * 5/9`", "_____no_output_____" ] ], [ [ "def fahr_to_celsius(temp):\n celsius = (temp - 32) * 5/9\n return celsius\n ", "_____no_output_____" ] ], [ [ "Now we can call the function `fahr_to_celsius` to convert a temperature from celsius to fahrenheit.\n\n", "_____no_output_____" ] ], [ [ "body_temp_f = 98.6\nbody_temp_c = fahr_to_celsius(body_temp_f)\nprint('Patient body temperature is:', body_temp_c, 'celsius')", "_____no_output_____" ] ], [ [ "## Reusing code with libraries", "_____no_output_____" ], [ "Python is a popular language for data analysis, so thankfully we can benefit from the hard work of others with the use of libraries. Pandas, a popular library for data analysis, introduces the `DataFrame`, a convenient data structure similar to a spreadsheet. Before using a library, we will need to import it.", "_____no_output_____" ] ], [ [ "# let's assign pandas an alias, pd, for brevity\nimport pandas as pd", "_____no_output_____" ] ], [ [ "We have shared a demo dataset online containing physiological data relating to 1000 patients admitted to an intensive care unit in Boston, Massachussetts, USA. Let's load this data into our new data structure.\n", "_____no_output_____" ] ], [ [ "url=\"https://raw.githubusercontent.com/tompollard/tableone/master/data/pn2012_demo.csv\"\ndata=pd.read_csv(url)", "_____no_output_____" ] ], [ [ "The variable `data` should now contain our new dataset. Let's view the first few rows using `head()`. Note: parentheses `\"()\"` are generally required when we are performing an action/operation. In this case, the action is to select a limited number of rows.", "_____no_output_____" ] ], [ [ "data.head()", "_____no_output_____" ] ], [ [ "We can perform other operations on the dataframe. For example, using `mean()` to get an average of the columns.", "_____no_output_____" ] ], [ [ "data.mean()", "_____no_output_____" ] ], [ [ "If we are unsure of the meaning of a method, we can check by adding `?` after the method. For example, what is `max`?", "_____no_output_____" ] ], [ [ "data.max?", "_____no_output_____" ], [ "data.max()", "_____no_output_____" ] ], [ [ "We can access a single column in the data by specifying the column name after the variable. For example, we can select a list of ages with `data.Age`, and then find the mean for this column in a similar way to before.", "_____no_output_____" ] ], [ [ "print('The mean age of patients is:', data.Age.mean())", "_____no_output_____" ] ], [ [ "Pandas also provides a convenient method `plot` for plotting data. Let's plot a distribution of the patient ages in our dataset.", "_____no_output_____" ] ], [ [ "data.Age.plot(kind='kde', title='Age of patients in years')", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import pandas as pd", "_____no_output_____" ], [ "df = pd.read_csv('data/survey_results_public.csv', index_col='Respondent')\nschema_df = pd.read_csv('data/survey_results_schema.csv', index_col='Column')", "_____no_output_____" ], [ "pd.set_option('display.max_columns', 85)\npd.set_option('display.max_rows', 85)", "_____no_output_____" ], [ "df.head()", "_____no_output_____" ], [ "df.sort_values(by=['Country', 'ConvertedComp'], ascending=[True, False], inplace=True)", "_____no_output_____" ], [ "df[['Country', 'ConvertedComp']].head(50)", "_____no_output_____" ], [ "df['ConvertedComp'].nlargest(10)", "_____no_output_____" ], [ "df.nsmallest(10, 'ConvertedComp')", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "<a href=\"https://colab.research.google.com/github/achmfirmansyah/sweet_project/blob/master/ICST2020/03_BUildup_transformation.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>", "_____no_output_____" ] ], [ [ "import pandas as pd\n!pip install rasterio\nimport rasterio\n!pip install geopandas\nimport geopandas as gpd\nimport numpy as np\nfrom google.colab import drive\ndrive.mount('/content/gdrive')\nimport numpy as np\nfrom sklearn.model_selection import train_test_split\n!pip install xgboost\nfrom xgboost import XGBClassifier\nfrom sklearn.metrics import classification_report, confusion_matrix", "Collecting rasterio\n\u001b[?25l Downloading https://files.pythonhosted.org/packages/02/7e/eed7dfd109fc89ed3cf8b5ed3f26f841b03b92f6ca1c31c4745f938a081b/rasterio-1.1.5-cp36-cp36m-manylinux1_x86_64.whl (18.2MB)\n\u001b[K |████████████████████████████████| 18.2MB 1.3MB/s \n\u001b[?25hCollecting affine\n Downloading https://files.pythonhosted.org/packages/ac/a6/1a39a1ede71210e3ddaf623982b06ecfc5c5c03741ae659073159184cd3e/affine-2.3.0-py2.py3-none-any.whl\nRequirement already satisfied: numpy in /usr/local/lib/python3.6/dist-packages (from rasterio) (1.18.5)\nRequirement already satisfied: click<8,>=4.0 in /usr/local/lib/python3.6/dist-packages (from rasterio) (7.1.2)\nRequirement already satisfied: attrs in /usr/local/lib/python3.6/dist-packages (from rasterio) (19.3.0)\nCollecting snuggs>=1.4.1\n Downloading https://files.pythonhosted.org/packages/cc/0e/d27d6e806d6c0d1a2cfdc5d1f088e42339a0a54a09c3343f7f81ec8947ea/snuggs-1.4.7-py3-none-any.whl\nCollecting cligj>=0.5\n Downloading https://files.pythonhosted.org/packages/e4/be/30a58b4b0733850280d01f8bd132591b4668ed5c7046761098d665ac2174/cligj-0.5.0-py3-none-any.whl\nCollecting click-plugins\n Downloading https://files.pythonhosted.org/packages/e9/da/824b92d9942f4e472702488857914bdd50f73021efea15b4cad9aca8ecef/click_plugins-1.1.1-py2.py3-none-any.whl\nRequirement already satisfied: pyparsing>=2.1.6 in /usr/local/lib/python3.6/dist-packages (from snuggs>=1.4.1->rasterio) (2.4.7)\nInstalling collected packages: affine, snuggs, cligj, click-plugins, rasterio\nSuccessfully installed affine-2.3.0 click-plugins-1.1.1 cligj-0.5.0 rasterio-1.1.5 snuggs-1.4.7\nCollecting geopandas\n\u001b[?25l Downloading https://files.pythonhosted.org/packages/f8/dd/c0a6429cc7692efd5c99420c9df525c40f472b50705871a770449027e244/geopandas-0.8.0-py2.py3-none-any.whl (962kB)\n\u001b[K |████████████████████████████████| 962kB 2.8MB/s \n\u001b[?25hCollecting pyproj>=2.2.0\n\u001b[?25l Downloading https://files.pythonhosted.org/packages/e5/c3/071e080230ac4b6c64f1a2e2f9161c9737a2bc7b683d2c90b024825000c0/pyproj-2.6.1.post1-cp36-cp36m-manylinux2010_x86_64.whl (10.9MB)\n\u001b[K |████████████████████████████████| 10.9MB 197kB/s \n\u001b[?25hCollecting fiona\n\u001b[?25l Downloading https://files.pythonhosted.org/packages/ec/20/4e63bc5c6e62df889297b382c3ccd4a7a488b00946aaaf81a118158c6f09/Fiona-1.8.13.post1-cp36-cp36m-manylinux1_x86_64.whl (14.7MB)\n\u001b[K |████████████████████████████████| 14.7MB 299kB/s \n\u001b[?25hRequirement already satisfied: shapely in /usr/local/lib/python3.6/dist-packages (from geopandas) (1.7.0)\nRequirement already satisfied: pandas>=0.23.0 in /usr/local/lib/python3.6/dist-packages (from geopandas) (1.0.5)\nCollecting munch\n Downloading https://files.pythonhosted.org/packages/cc/ab/85d8da5c9a45e072301beb37ad7f833cd344e04c817d97e0cc75681d248f/munch-2.5.0-py2.py3-none-any.whl\nRequirement already satisfied: click<8,>=4.0 in /usr/local/lib/python3.6/dist-packages (from fiona->geopandas) (7.1.2)\nRequirement already satisfied: cligj>=0.5 in /usr/local/lib/python3.6/dist-packages (from fiona->geopandas) (0.5.0)\nRequirement already satisfied: click-plugins>=1.0 in /usr/local/lib/python3.6/dist-packages (from fiona->geopandas) (1.1.1)\nRequirement already satisfied: six>=1.7 in /usr/local/lib/python3.6/dist-packages (from fiona->geopandas) (1.12.0)\nRequirement already satisfied: attrs>=17 in /usr/local/lib/python3.6/dist-packages (from fiona->geopandas) (19.3.0)\nRequirement already satisfied: numpy>=1.13.3 in /usr/local/lib/python3.6/dist-packages (from pandas>=0.23.0->geopandas) (1.18.5)\nRequirement already satisfied: pytz>=2017.2 in /usr/local/lib/python3.6/dist-packages (from pandas>=0.23.0->geopandas) (2018.9)\nRequirement already satisfied: python-dateutil>=2.6.1 in /usr/local/lib/python3.6/dist-packages (from pandas>=0.23.0->geopandas) (2.8.1)\nInstalling collected packages: pyproj, munch, fiona, geopandas\nSuccessfully installed fiona-1.8.13.post1 geopandas-0.8.0 munch-2.5.0 pyproj-2.6.1.post1\nGo to this URL in a browser: https://accounts.google.com/o/oauth2/auth?client_id=947318989803-6bn6qk8qdgf4n4g3pfee6491hc0brc4i.apps.googleusercontent.com&redirect_uri=urn%3aietf%3awg%3aoauth%3a2.0%3aoob&response_type=code&scope=email%20https%3a%2f%2fwww.googleapis.com%2fauth%2fdocs.test%20https%3a%2f%2fwww.googleapis.com%2fauth%2fdrive%20https%3a%2f%2fwww.googleapis.com%2fauth%2fdrive.photos.readonly%20https%3a%2f%2fwww.googleapis.com%2fauth%2fpeopleapi.readonly\n\nEnter your authorization code:\n··········\nMounted at /content/gdrive\nRequirement already satisfied: xgboost in /usr/local/lib/python3.6/dist-packages (0.90)\nRequirement already satisfied: numpy in /usr/local/lib/python3.6/dist-packages (from xgboost) (1.18.5)\nRequirement already satisfied: scipy in /usr/local/lib/python3.6/dist-packages (from xgboost) (1.4.1)\n" ], [ "lokus=['jabodetabek','mebidangro','maminasata','kedungsepur']\nconfussion_matrix_list=[]\nclassification_report_list=[]", "_____no_output_____" ], [ "#Create model\nfor lokasi in lokus:\n print(lokasi)\n dataset = rasterio.open('/content/gdrive/My Drive/Urban_monitoring/Urban_/compiled_GHSL_30_train_2014_'+lokasi+'.tif')\n dataset_bands=pd.DataFrame()\n for i in dataset.indexes:\n temp=dataset.read(i)\n temp=pd.DataFrame(data=np.array(temp).flatten()).rename(columns={0:i})\n dataset_bands=temp.join(dataset_bands)\n dataset_bands.rename(columns={1:'BU_class',2:'NDVI_landsat7',3:'NDBI_landsat7',4:'MNDWI_landsat7',\n 5:'SAVI_landsat7',6:'LSE_landsat7',7:'rad_VIIRS'},inplace=True)\n dataset_bands['U_class']=dataset_bands.BU_class.apply(lambda y: 1 if y>=3 else 0)\n dataset_bands['LSE_der_landsat7']=dataset_bands.NDVI_landsat7.apply(lambda y: 0.995 if y < -0.185 else (\n 0.970 if y< 0.157 else (1.0098+0.047*np.log(y) if y<0.727 else 0.990)))\n dataset=dataset_bands.query('U_class==0').sample(10000).append(dataset_bands.query('U_class==1').sample(10000))\n dataset=dataset.reset_index()[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDBI_landsat7','NDVI_landsat7','LSE_der_landsat7','U_class']]\n X_train, X_test, y_train, y_test = train_test_split(dataset[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDVI_landsat7','NDBI_landsat7','LSE_der_landsat7']],\n dataset[['U_class']], test_size = 0.20)\n xgclassifier = XGBClassifier(random_state=123,colsample_bytree= 0.7, \n learning_rate=0.05, max_depth= 4, \n min_child_weight=11, n_estimators= 500, nthread= 4, objective= 'binary:logistic', \n seed= 123, silent=1, subsample= 0.8)\n xgclassifier.fit(X_train.values,y_train.values)\n y_pred=xgclassifier.predict(X_test.values)\n #confussion_matrix_list.append(confusion_matrix(y_test, y_pred))\n classification_report_list.append(classification_report(y_test, y_pred))\n dataset=dataset_bands.reset_index()[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDBI_landsat7','NDVI_landsat7','LSE_der_landsat7']]\n class_2014_=xgclassifier.predict(dataset.values)\n confussion_matrix_list.append(confusion_matrix(dataset_bands[['U_class']],class_2014_))\n temp_ = rasterio.open('/content/gdrive/My Drive/Urban_monitoring/Urban_/compiled_GHSL_30_train_2014_'+lokasi+'.tif')\n temp = temp_.read(1)\n \n array_class_2014=class_2014_.reshape(temp.shape[0],temp.shape[1]).astype(np.uint8)\n \n profile = temp_.profile\n profile.update(\n dtype=array_class_2014.dtype,\n count=1,\n compress='lzw')\n with rasterio.open(\n '/content/gdrive/My Drive/Urban_monitoring/Urban_/GHSL_/GHSL_rev/rev_class_2014_'+lokasi+'.tif','w',**profile) as dst2:\n dst2.write(array_class_2014,1)\n dst2.close()", "jabodetabek\n" ], [ "confussion_matrix_list[3]", "_____no_output_____" ], [ "#Create model\nfor lokasi in lokus:\n print(lokasi)\n dataset = rasterio.open('/content/gdrive/My Drive/Urban_monitoring/Urban_/compiled_GHSL_30_train_2014_'+lokasi+'.tif')\n dataset_bands=pd.DataFrame()\n for i in dataset.indexes:\n temp=dataset.read(i)\n temp=pd.DataFrame(data=np.array(temp).flatten()).rename(columns={0:i})\n dataset_bands=temp.join(dataset_bands)\n dataset_bands.rename(columns={1:'BU_class',2:'NDVI_landsat7',3:'NDBI_landsat7',4:'MNDWI_landsat7',\n 5:'SAVI_landsat7',6:'LSE_landsat7',7:'rad_VIIRS'},inplace=True)\n dataset_bands['U_class']=dataset_bands.BU_class.apply(lambda y: 1 if y>=3 else 0)\n dataset_bands['LSE_der_landsat7']=dataset_bands.NDVI_landsat7.apply(lambda y: 0.995 if y < -0.185 else (\n 0.970 if y< 0.157 else (1.0098+0.047*np.log(y) if y<0.727 else 0.990)))\n dataset=dataset_bands.query('U_class==0').sample(10000).append(dataset_bands.query('U_class==1').sample(10000))\n dataset=dataset.reset_index()[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDBI_landsat7','NDVI_landsat7','LSE_der_landsat7','U_class']]\n X_train, X_test, y_train, y_test = train_test_split(dataset[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDVI_landsat7','NDBI_landsat7','LSE_der_landsat7']],\n dataset[['U_class']], test_size = 0.20)\n xgclassifier = XGBClassifier(random_state=123,colsample_bytree= 0.7, \n learning_rate=0.05, max_depth= 4, \n min_child_weight=11, n_estimators= 500, nthread= 4, objective= 'binary:logistic', \n seed= 123, silent=1, subsample= 0.8)\n xgclassifier.fit(X_train.values,y_train.values)\n y_pred=xgclassifier.predict(X_test.values)\n confussion_matrix_list.append(confusion_matrix(y_test, y_pred))\n classification_report_list.append(classification_report(y_test, y_pred))\n\n #classification_2019:\n #dataset = rasterio.open('/content/gdrive/My Drive/Urban_monitoring/Urban_/compiled_GHSL_30_train_2019_'+lokasi+'.tif')\n #dataset_bands=pd.DataFrame()\n #for i in dataset.indexes:\n # temp=dataset.read(i)\n # temp=pd.DataFrame(data=np.array(temp).flatten()).rename(columns={0:i})\n # dataset_bands=temp.join(dataset_bands)\n #dataset_bands.rename(columns={1:'BU_class',2:'NDVI_landsat7',3:'NDBI_landsat7',4:'MNDWI_landsat7',\n # 5:'SAVI_landsat7',6:'LSE_landsat7',7:'rad_VIIRS'},inplace=True)\n #dataset_bands['U_class']=dataset_bands.BU_class.apply(lambda y: 1 if y>=3 else 0)\n #dataset_bands['LSE_der_landsat7']=dataset_bands.NDVI_landsat7.apply(lambda y: 0.995 if y < -0.185 else (\n # 0.970 if y< 0.157 else (1.0098+0.047*np.log(y) if y<0.727 else 0.990)))\n #dataset=dataset_bands.reset_index()[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDBI_landsat7','NDVI_landsat7','LSE_der_landsat7']]\n #class_2019_=xgclassifier.predict(dataset.values)\n #temp_ = rasterio.open('/content/gdrive/My Drive/Urban_monitoring/Urban_/compiled_GHSL_30_train_2019_'+lokasi+'.tif')\n #temp = temp_.read(1)\n #array_class_2019=class_2019_.reshape(temp.shape[0],temp.shape[1]).astype(np.uint8)\n #profile = temp_.profile\n #profile.update(\n # dtype=array_class_2019.dtype,\n # count=1,\n # compress='lzw')\n #with rasterio.open(\n # '/content/gdrive/My Drive/Urban_monitoring/Urban_/GHSL_/GHSL_rev/rev_class_2019_'+lokasi+'.tif','w',**profile) as dst2:\n # dst2.write(array_class_2019,1)\n # dst2.close()", "jabodetabek\n" ], [ "#Create model\nfor lokasi in lokus:\n #print(lokasi)\n #dataset = rasterio.open('/content/gdrive/My Drive/Urban_monitoring/Urban_/compiled_GHSL_30_train_2014_'+lokasi+'.tif')\n #dataset_bands=pd.DataFrame()\n #for i in dataset.indexes:\n # temp=dataset.read(i)\n # temp=pd.DataFrame(data=np.array(temp).flatten()).rename(columns={0:i})\n # dataset_bands=temp.join(dataset_bands)\n #dataset_bands.rename(columns={1:'BU_class',2:'NDVI_landsat7',3:'NDBI_landsat7',4:'MNDWI_landsat7',\n # 5:'SAVI_landsat7',6:'LSE_landsat7',7:'rad_VIIRS'},inplace=True)\n #dataset_bands['U_class']=dataset_bands.BU_class.apply(lambda y: 1 if y>=3 else 0)\n #dataset_bands['LSE_der_landsat7']=dataset_bands.NDVI_landsat7.apply(lambda y: 0.995 if y < -0.185 else (\n # 0.970 if y< 0.157 else (1.0098+0.047*np.log(y) if y<0.727 else 0.990)))\n #dataset=dataset_bands.query('U_class==0').sample(10000).append(dataset_bands.query('U_class==1').sample(10000))\n #dataset=dataset.reset_index()[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDBI_landsat7','NDVI_landsat7','LSE_der_landsat7','U_class']]\n #X_train, X_test, y_train, y_test = train_test_split(dataset[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDVI_landsat7','NDBI_landsat7','LSE_der_landsat7']],\n # dataset[['U_class']], test_size = 0.20)\n #xgclassifier = XGBClassifier(random_state=123,colsample_bytree= 0.7, \n # learning_rate=0.05, max_depth= 4, \n # min_child_weight=11, n_estimators= 500, nthread= 4, objective= 'binary:logistic', \n # seed= 123, silent=1, subsample= 0.8)\n #xgclassifier.fit(X_train.values,y_train.values)\n #y_pred=xgclassifier.predict(X_test.values)\n\n #classification_2015:\n #dataset = rasterio.open('/content/gdrive/My Drive/Urban_monitoring/Urban_/compiled_GHSL_30_train_2015_'+lokasi+'.tif')\n #dataset_bands=pd.DataFrame()\n #for i in dataset.indexes:\n # temp=dataset.read(i)\n # temp=pd.DataFrame(data=np.array(temp).flatten()).rename(columns={0:i})\n # dataset_bands=temp.join(dataset_bands)\n #dataset_bands.rename(columns={1:'BU_class',2:'NDVI_landsat7',3:'NDBI_landsat7',4:'MNDWI_landsat7',\n # 5:'SAVI_landsat7',6:'LSE_landsat7',7:'rad_VIIRS'},inplace=True)\n #dataset_bands['U_class']=dataset_bands.BU_class.apply(lambda y: 1 if y>=3 else 0)\n #dataset_bands['LSE_der_landsat7']=dataset_bands.NDVI_landsat7.apply(lambda y: 0.995 if y < -0.185 else (\n # 0.970 if y< 0.157 else (1.0098+0.047*np.log(y) if y<0.727 else 0.990)))\n #dataset=dataset_bands.reset_index()[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDBI_landsat7','NDVI_landsat7','LSE_der_landsat7']]\n #class_2015_=xgclassifier.predict(dataset.values)\n #temp_ = rasterio.open('/content/gdrive/My Drive/Urban_monitoring/Urban_/compiled_GHSL_30_train_2015_'+lokasi+'.tif')\n #temp = temp_.read(1)\n #array_class_2015=class_2015_.reshape(temp.shape[0],temp.shape[1]).astype(np.uint8)\n #profile = temp_.profile\n #profile.update(\n # dtype=array_class_2015.dtype,\n # count=1,\n # compress='lzw')\n #with rasterio.open(\n # '/content/gdrive/My Drive/Urban_monitoring/Urban_/GHSL_/GHSL_rev/rev_class_2015_'+lokasi+'.tif','w',**profile) as dst2:\n # dst2.write(array_class_2015,1)\n # dst2.close()", "jabodetabek\n" ], [ "lokus=['gerbangkertasusila']\n#Create model\nfor lokasi in lokus:\n print(lokasi)\n dataset = rasterio.open('/content/gdrive/My Drive/Urban_monitoring/Urban_/compiled_GHSL_30_train_2014_'+lokasi+'.tif')\n dataset_bands=pd.DataFrame()\n for i in dataset.indexes:\n temp=dataset.read(i)\n temp=pd.DataFrame(data=np.array(temp).flatten()).rename(columns={0:i})\n dataset_bands=temp.join(dataset_bands)\n dataset_bands.rename(columns={1:'BU_class',2:'NDVI_landsat7',3:'NDBI_landsat7',4:'MNDWI_landsat7',\n 5:'SAVI_landsat7',6:'LSE_landsat7',7:'rad_VIIRS'},inplace=True)\n dataset_bands['U_class']=dataset_bands.BU_class.apply(lambda y: 1 if y>=3 else 0)\n dataset_bands['LSE_der_landsat7']=dataset_bands.NDVI_landsat7.apply(lambda y: 0.995 if y < -0.185 else (\n 0.970 if y< 0.157 else (1.0098+0.047*np.log(y) if y<0.727 else 0.990)))\n dataset=dataset_bands.query('U_class==0').sample(10000).append(dataset_bands.query('U_class==1').sample(10000))\n dataset=dataset.reset_index()[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDBI_landsat7','NDVI_landsat7','LSE_der_landsat7','U_class']]\n X_train, X_test, y_train, y_test = train_test_split(dataset[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDVI_landsat7','NDBI_landsat7','LSE_der_landsat7']],\n dataset[['U_class']], test_size = 0.20)\n xgclassifier = XGBClassifier(random_state=123,colsample_bytree= 0.7, \n learning_rate=0.05, max_depth= 5, \n min_child_weight=11, n_estimators= 500, nthread= 4, objective= 'binary:logistic', \n seed= 123, silent=1, subsample= 0.8)\n xgclassifier.fit(X_train.values,y_train.values)\n #y_pred=xgclassifier.predict(X_test.values)\n y_pred=xgclassifier.predict(dataset_bands[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDVI_landsat7','NDBI_landsat7','LSE_der_landsat7']].values)\n confussion_matrix_list.append(confusion_matrix(dataset_bands[['U_class']],y_pred))\n #confussion_matrix_list.append(confusion_matrix(y_test, y_pred))\n #classification_report_list.append(classification_report(y_test, y_pred))\n\n #classification_2015:\n #dataset = rasterio.open('/content/gdrive/My Drive/Urban_monitoring/Urban_/compiled_GHSL_30_train_2015_'+lokasi+'.tif')\n #dataset_bands=pd.DataFrame()\n #for i in dataset.indexes:\n # temp=dataset.read(i)\n # temp=pd.DataFrame(data=np.array(temp).flatten()).rename(columns={0:i})\n # dataset_bands=temp.join(dataset_bands)\n #dataset_bands.rename(columns={1:'BU_class',2:'NDVI_landsat7',3:'NDBI_landsat7',4:'MNDWI_landsat7',\n # 5:'SAVI_landsat7',6:'LSE_landsat7',7:'rad_VIIRS'},inplace=True)\n #dataset_bands['U_class']=dataset_bands.BU_class.apply(lambda y: 1 if y>=3 else 0)\n #dataset_bands['LSE_der_landsat7']=dataset_bands.NDVI_landsat7.apply(lambda y: 0.995 if y < -0.185 else (\n # 0.970 if y< 0.157 else (1.0098+0.047*np.log(y) if y<0.727 else 0.990)))\n #dataset=dataset_bands.reset_index()[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDBI_landsat7','NDVI_landsat7','LSE_der_landsat7']]\n #class_2015_=xgclassifier.predict(dataset.values)\n #temp_ = rasterio.open('/content/gdrive/My Drive/Urban_monitoring/Urban_/compiled_GHSL_30_train_2015_'+lokasi+'.tif')\n #temp = temp_.read(1)\n #array_class_2015=class_2015_.reshape(temp.shape[0],temp.shape[1]).astype(np.uint8)\n #profile = temp_.profile\n #profile.update(\n # dtype=array_class_2015.dtype,\n # count=1,\n # compress='lzw')\n #with rasterio.open(\n # '/content/gdrive/My Drive/Urban_monitoring/Urban_/GHSL_/GHSL_rev/rev_class_2015_'+lokasi+'.tif','w',**profile) as dst2:\n # dst2.write(array_class_2015,1)\n # dst2.close()\n#Create model\n#for lokasi in lokus:\n # print(lokasi)\n #dataset = rasterio.open('/content/gdrive/My Drive/Urban_monitoring/Urban_/compiled_GHSL_30_train_2014_'+lokasi+'.tif')\n #dataset_bands=pd.DataFrame()\n #for i in dataset.indexes:\n # temp=dataset.read(i)\n # temp=pd.DataFrame(data=np.array(temp).flatten()).rename(columns={0:i})\n # dataset_bands=temp.join(dataset_bands)\n #dataset_bands.rename(columns={1:'BU_class',2:'NDVI_landsat7',3:'NDBI_landsat7',4:'MNDWI_landsat7',\n # 5:'SAVI_landsat7',6:'LSE_landsat7',7:'rad_VIIRS'},inplace=True)\n #dataset_bands['U_class']=dataset_bands.BU_class.apply(lambda y: 1 if y>=3 else 0)\n #dataset_bands['LSE_der_landsat7']=dataset_bands.NDVI_landsat7.apply(lambda y: 0.995 if y < -0.185 else (\n # 0.970 if y< 0.157 else (1.0098+0.047*np.log(y) if y<0.727 else 0.990)))\n #dataset=dataset_bands.query('U_class==0').sample(10000).append(dataset_bands.query('U_class==1').sample(10000))\n #dataset=dataset.reset_index()[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDBI_landsat7','NDVI_landsat7','LSE_der_landsat7','U_class']]\n #X_train, X_test, y_train, y_test = train_test_split(dataset[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDVI_landsat7','NDBI_landsat7','LSE_der_landsat7']],\n # dataset[['U_class']], test_size = 0.20)\n #xgclassifier = XGBClassifier(random_state=123,colsample_bytree= 0.7, \n # learning_rate=0.05, max_depth= 5, \n # min_child_weight=11, n_estimators= 500, nthread= 4, objective= 'binary:logistic', \n # seed= 123, silent=1, subsample= 0.8)\n #xgclassifier.fit(X_train.values,y_train.values)\n #y_pred=xgclassifier.predict(X_test.values)\n #confussion_matrix_list.append(confusion_matrix(y_test, y_pred))\n #classification_report_list.append(classification_report(y_test, y_pred))\n\n #classification_2019:\n #dataset = rasterio.open('/content/gdrive/My Drive/Urban_monitoring/Urban_/compiled_GHSL_30_train_2019_'+lokasi+'.tif')\n #dataset_bands=pd.DataFrame()\n #for i in dataset.indexes:\n # temp=dataset.read(i)\n # temp=pd.DataFrame(data=np.array(temp).flatten()).rename(columns={0:i})\n # dataset_bands=temp.join(dataset_bands)\n #dataset_bands.rename(columns={1:'BU_class',2:'NDVI_landsat7',3:'NDBI_landsat7',4:'MNDWI_landsat7',\n # 5:'SAVI_landsat7',6:'LSE_landsat7',7:'rad_VIIRS'},inplace=True)\n #dataset_bands['U_class']=dataset_bands.BU_class.apply(lambda y: 1 if y>=3 else 0)\n #dataset_bands['LSE_der_landsat7']=dataset_bands.NDVI_landsat7.apply(lambda y: 0.995 if y < -0.185 else (\n # 0.970 if y< 0.157 else (1.0098+0.047*np.log(y) if y<0.727 else 0.990)))\n #dataset=dataset_bands.reset_index()[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDBI_landsat7','NDVI_landsat7','LSE_der_landsat7']]\n #class_2019_=xgclassifier.predict(dataset.values)\n #temp_ = rasterio.open('/content/gdrive/My Drive/Urban_monitoring/Urban_/compiled_GHSL_30_train_2019_'+lokasi+'.tif')\n #temp = temp_.read(1)\n #array_class_2019=class_2019_.reshape(temp.shape[0],temp.shape[1]).astype(np.uint8)\n #profile = temp_.profile\n #profile.update(\n # dtype=array_class_2019.dtype,\n # count=1,\n # compress='lzw')\n #with rasterio.open(\n # '/content/gdrive/My Drive/Urban_monitoring/Urban_/GHSL_/GHSL_rev/rev_class_2019_'+lokasi+'.tif','w',**profile) as dst2:\n # dst2.write(array_class_2019,1)\n # dst2.close()\n\n#Create model\n#for lokasi in lokus:\n #print(lokasi)\n #dataset = rasterio.open('/content/gdrive/My Drive/Urban_monitoring/Urban_/compiled_GHSL_30_train_2014_'+lokasi+'.tif')\n #dataset_bands=pd.DataFrame()\n #for i in dataset.indexes:\n # temp=dataset.read(i)\n # temp=pd.DataFrame(data=np.array(temp).flatten()).rename(columns={0:i})\n # dataset_bands=temp.join(dataset_bands)\n #dataset_bands.rename(columns={1:'BU_class',2:'NDVI_landsat7',3:'NDBI_landsat7',4:'MNDWI_landsat7',\n # 5:'SAVI_landsat7',6:'LSE_landsat7',7:'rad_VIIRS'},inplace=True)\n #dataset_bands['U_class']=dataset_bands.BU_class.apply(lambda y: 1 if y>=3 else 0)\n #dataset_bands['LSE_der_landsat7']=dataset_bands.NDVI_landsat7.apply(lambda y: 0.995 if y < -0.185 else (\n # 0.970 if y< 0.157 else (1.0098+0.047*np.log(y) if y<0.727 else 0.990)))\n #dataset=dataset_bands.query('U_class==0').sample(10000).append(dataset_bands.query('U_class==1').sample(10000))\n #dataset=dataset.reset_index()[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDBI_landsat7','NDVI_landsat7','LSE_der_landsat7','U_class']]\n #X_train, X_test, y_train, y_test = train_test_split(dataset[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDVI_landsat7','NDBI_landsat7','LSE_der_landsat7']],\n # dataset[['U_class']], test_size = 0.20)\n #xgclassifier = XGBClassifier(random_state=123,colsample_bytree= 0.7, \n # learning_rate=0.05, max_depth= 5, \n # min_child_weight=11, n_estimators= 500, nthread= 4, objective= 'binary:logistic', \n # seed= 123, silent=1, subsample= 0.8)\n #xgclassifier.fit(X_train.values,y_train.values)\n #y_pred=xgclassifier.predict(X_test.values)\n #confussion_matrix_list.append(confusion_matrix(y_test, y_pred))\n #classification_report_list.append(classification_report(y_test, y_pred))\n\n #classification_2019:\n #dataset = rasterio.open('/content/gdrive/My Drive/Urban_monitoring/Urban_/compiled_GHSL_30_train_2014_'+lokasi+'.tif')\n #dataset_bands=pd.DataFrame()\n #for i in dataset.indexes:\n # temp=dataset.read(i)\n # temp=pd.DataFrame(data=np.array(temp).flatten()).rename(columns={0:i})\n # dataset_bands=temp.join(dataset_bands)\n #dataset_bands.rename(columns={1:'BU_class',2:'NDVI_landsat7',3:'NDBI_landsat7',4:'MNDWI_landsat7',\n # 5:'SAVI_landsat7',6:'LSE_landsat7',7:'rad_VIIRS'},inplace=True)\n #dataset_bands['U_class']=dataset_bands.BU_class.apply(lambda y: 1 if y>=3 else 0)\n #dataset_bands['LSE_der_landsat7']=dataset_bands.NDVI_landsat7.apply(lambda y: 0.995 if y < -0.185 else (\n # 0.970 if y< 0.157 else (1.0098+0.047*np.log(y) if y<0.727 else 0.990)))\n #dataset=dataset_bands.reset_index()[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDBI_landsat7','NDVI_landsat7','LSE_der_landsat7']]\n #class_2014_=xgclassifier.predict(dataset.values)\n #temp_ = rasterio.open('/content/gdrive/My Drive/Urban_monitoring/Urban_/compiled_GHSL_30_train_2014_'+lokasi+'.tif')\n #temp = temp_.read(1)\n #array_class_2014=class_2014_.reshape(temp.shape[0],temp.shape[1]).astype(np.uint8)\n #profile = temp_.profile\n #profile.update(\n # dtype=array_class_2014.dtype,\n # count=1,\n # compress='lzw')\n #with rasterio.open(\n # '/content/gdrive/My Drive/Urban_monitoring/Urban_/GHSL_/GHSL_rev/rev_class_2014_'+lokasi+'.tif','w',**profile) as dst2:\n # dst2.write(array_class_2014,1)\n # dst2.close()", "gerbangkertasusila\n" ], [ "confussion_matrix_list[4]", "_____no_output_____" ], [ "lokus=['bandungraya']\nfrom sklearn.ensemble import RandomForestClassifier\n#Create model\nfor lokasi in lokus:\n print(lokasi)\n dataset = rasterio.open('/content/gdrive/My Drive/Urban_monitoring/Urban_/compiled_GHSL_30_train_2014_'+lokasi+'.tif')\n dataset_bands=pd.DataFrame()\n for i in dataset.indexes:\n temp=dataset.read(i)\n temp=pd.DataFrame(data=np.array(temp).flatten()).rename(columns={0:i})\n dataset_bands=temp.join(dataset_bands)\n dataset_bands.rename(columns={1:'BU_class',2:'NDVI_landsat7',3:'NDBI_landsat7',4:'MNDWI_landsat7',\n 5:'SAVI_landsat7',6:'LSE_landsat7',7:'rad_VIIRS'},inplace=True)\n dataset_bands['U_class']=dataset_bands.BU_class.apply(lambda y: 1 if y>=3 else 0)\n dataset_bands['LSE_der_landsat7']=dataset_bands.NDVI_landsat7.apply(lambda y: 0.995 if y < -0.185 else (\n 0.970 if y< 0.157 else (1.0098+0.047*np.log(y) if y<0.727 else 0.990)))\n dataset=dataset_bands.query('U_class==0').sample(10000).append(dataset_bands.query('U_class==1').sample(10000))\n dataset=dataset.reset_index()[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDBI_landsat7','NDVI_landsat7','LSE_der_landsat7','U_class']]\n X_train, X_test, y_train, y_test = train_test_split(dataset[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDVI_landsat7','NDBI_landsat7','LSE_der_landsat7']],\n dataset[['U_class']], test_size = 0.20)\n rfclassifier = RandomForestClassifier(random_state=123,max_depth=6,n_estimators=500,criterion='entropy')\n rfclassifier.fit(X_train.values,y_train.values)\n y_pred=rfclassifier.predict(dataset_bands[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDBI_landsat7','NDVI_landsat7','LSE_der_landsat7']].values)\n confussion_matrix_list.append(confusion_matrix(dataset_bands['U_class'],y_pred))\n #y_pred=rfclassifier.predict(X_test.values)\n\n #classification_2014:\n #dataset = rasterio.open('/content/gdrive/My Drive/Urban_monitoring/Urban_/compiled_GHSL_30_train_2015_'+lokasi+'.tif')\n #dataset_bands=pd.DataFrame()\n #for i in dataset.indexes:\n # temp=dataset.read(i)\n # temp=pd.DataFrame(data=np.array(temp).flatten()).rename(columns={0:i})\n # dataset_bands=temp.join(dataset_bands)\n #dataset_bands.rename(columns={1:'BU_class',2:'NDVI_landsat7',3:'NDBI_landsat7',4:'MNDWI_landsat7',\n # 5:'SAVI_landsat7',6:'LSE_landsat7',7:'rad_VIIRS'},inplace=True)\n #dataset_bands['U_class']=dataset_bands.BU_class.apply(lambda y: 1 if y>=3 else 0)\n ##dataset_bands['LSE_der_landsat7']=dataset_bands.NDVI_landsat7.apply(lambda y: 0.995 if y < -0.185 else (\n # 0.970 if y< 0.157 else (1.0098+0.047*np.log(y) if y<0.727 else 0.990)))\n #dataset=dataset_bands.reset_index()[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDBI_landsat7','NDVI_landsat7','LSE_der_landsat7']]\n #class_2015_=rfclassifier.predict(dataset.values)\n #dataset_bands['U_class']\n #temp_ = rasterio.open('/content/gdrive/My Drive/Urban_monitoring/Urban_/compiled_GHSL_30_train_2015_'+lokasi+'.tif')\n #temp = temp_.read(1)\n #array_class_2015=class_2015_.reshape(temp.shape[0],temp.shape[1]).astype(np.uint8)\n #profile = temp_.profile\n #profile.update(\n # dtype=array_class_2015.dtype,\n # count=1,\n # compress='lzw')\n #with rasterio.open(\n # '/content/gdrive/My Drive/Urban_monitoring/Urban_/GHSL_/GHSL_rev/rev_class_2015_'+lokasi+'_2.tif','w',**profile) as dst2:\n # dst2.write(array_class_2015,1)\n # dst2.close()\n#Create model\n#for lokasi in lokus:\n # print(lokasi)\n # dataset = rasterio.open('/content/gdrive/My Drive/Urban_monitoring/Urban_/compiled_GHSL_30_train_2014_'+lokasi+'.tif')\n # dataset_bands=pd.DataFrame()\n # for i in dataset.indexes:\n # temp=dataset.read(i)\n # temp=pd.DataFrame(data=np.array(temp).flatten()).rename(columns={0:i})\n # dataset_bands=temp.join(dataset_bands)\n # dataset_bands.rename(columns={1:'BU_class',2:'NDVI_landsat7',3:'NDBI_landsat7',4:'MNDWI_landsat7',\n # 5:'SAVI_landsat7',6:'LSE_landsat7',7:'rad_VIIRS'},inplace=True)\n #dataset_bands['U_class']=dataset_bands.BU_class.apply(lambda y: 1 if y>=3 else 0)\n #dataset_bands['LSE_der_landsat7']=dataset_bands.NDVI_landsat7.apply(lambda y: 0.995 if y < -0.185 else (\n # 0.970 if y< 0.157 else (1.0098+0.047*np.log(y) if y<0.727 else 0.990)))\n #dataset=dataset_bands.query('U_class==0').sample(10000).append(dataset_bands.query('U_class==1').sample(10000))\n #dataset=dataset.reset_index()[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDBI_landsat7','NDVI_landsat7','LSE_der_landsat7','U_class']]\n #X_train, X_test, y_train, y_test = train_test_split(dataset[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDVI_landsat7','NDBI_landsat7','LSE_der_landsat7']],\n # dataset[['U_class']], test_size = 0.20)\n #rfclassifier = RandomForestClassifier(random_state=123,max_depth=6,n_estimators=500,criterion='entropy')\n #rfclassifier.fit(X_train.values,y_train.values)\n #y_pred=rfclassifier.predict(X_test.values)\n #confussion_matrix_list.append(confusion_matrix(y_test, y_pred))\n #classification_report_list.append(classification_report(y_test, y_pred))\n\n #classification_2019:\n #dataset = rasterio.open('/content/gdrive/My Drive/Urban_monitoring/Urban_/compiled_GHSL_30_train_2019_'+lokasi+'.tif')\n #dataset_bands=pd.DataFrame()\n #for i in dataset.indexes:\n # temp=dataset.read(i)\n # temp=pd.DataFrame(data=np.array(temp).flatten()).rename(columns={0:i})\n # dataset_bands=temp.join(dataset_bands)\n #dataset_bands.rename(columns={1:'BU_class',2:'NDVI_landsat7',3:'NDBI_landsat7',4:'MNDWI_landsat7',\n # 5:'SAVI_landsat7',6:'LSE_landsat7',7:'rad_VIIRS'},inplace=True)\n #dataset_bands['U_class']=dataset_bands.BU_class.apply(lambda y: 1 if y>=3 else 0)\n #dataset_bands['LSE_der_landsat7']=dataset_bands.NDVI_landsat7.apply(lambda y: 0.995 if y < -0.185 else (\n # 0.970 if y< 0.157 else (1.0098+0.047*np.log(y) if y<0.727 else 0.990)))\n #dataset=dataset_bands.reset_index()[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDBI_landsat7','NDVI_landsat7','LSE_der_landsat7']]\n #class_2019_=rfclassifier.predict(dataset.values)\n #temp_ = rasterio.open('/content/gdrive/My Drive/Urban_monitoring/Urban_/compiled_GHSL_30_train_2019_'+lokasi+'.tif')\n #temp = temp_.read(1)\n #array_class_2019=class_2019_.reshape(temp.shape[0],temp.shape[1]).astype(np.uint8)\n #profile = temp_.profile\n #profile.update(\n # dtype=array_class_2019.dtype,\n # count=1,\n # compress='lzw')\n #with rasterio.open(\n # '/content/gdrive/My Drive/Urban_monitoring/Urban_/GHSL_/GHSL_rev/rev_class_2019_'+lokasi+'.tif','w',**profile) as dst2:\n # dst2.write(array_class_2019,1)\n # dst2.close()\n\n#Create model\n#for lokasi in lokus:\n# print(lokasi)\n# dataset = rasterio.open('/content/gdrive/My Drive/Urban_monitoring/Urban_/compiled_GHSL_30_train_2014_'+lokasi+'.tif')\n# dataset_bands=pd.DataFrame()\n# for i in dataset.indexes:\n# temp=dataset.read(i)\n# temp=pd.DataFrame(data=np.array(temp).flatten()).rename(columns={0:i})\n# dataset_bands=temp.join(dataset_bands)\n# dataset_bands.rename(columns={1:'BU_class',2:'NDVI_landsat7',3:'NDBI_landsat7',4:'MNDWI_landsat7',\n# 5:'SAVI_landsat7',6:'LSE_landsat7',7:'rad_VIIRS'},inplace=True)\n# dataset_bands['U_class']=dataset_bands.BU_class.apply(lambda y: 1 if y>=3 else 0)\n# dataset_bands['LSE_der_landsat7']=dataset_bands.NDVI_landsat7.apply(lambda y: 0.995 if y < -0.185 else (\n# 0.970 if y< 0.157 else (1.0098+0.047*np.log(y) if y<0.727 else 0.990)))\n# dataset=dataset_bands.query('U_class==0').sample(10000).append(dataset_bands.query('U_class==1').sample(10000))\n# dataset=dataset.reset_index()[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDBI_landsat7','NDVI_landsat7','LSE_der_landsat7','U_class']]\n# X_train, X_test, y_train, y_test = train_test_split(dataset[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDVI_landsat7','NDBI_landsat7','LSE_der_landsat7']],\n# dataset[['U_class']], test_size = 0.20)\n# rfclassifier = RandomForestClassifier(random_state=123,max_depth=6,n_estimators=500,criterion='entropy')\n# rfclassifier.fit(X_train.values,y_train.values)\n# y_pred=rfclassifier.predict(X_test.values)\n# confussion_matrix_list.append(confusion_matrix(y_test, y_pred))\n# classification_report_list.append(classification_report(y_test, y_pred))\n\n #classification_2019:\n# dataset = rasterio.open('/content/gdrive/My Drive/Urban_monitoring/Urban_/compiled_GHSL_30_train_2014_'+lokasi+'.tif')\n# dataset_bands=pd.DataFrame()\n# for i in dataset.indexes:\n# temp=dataset.read(i)\n# temp=pd.DataFrame(data=np.array(temp).flatten()).rename(columns={0:i})\n# dataset_bands=temp.join(dataset_bands)\n# dataset_bands.rename(columns={1:'BU_class',2:'NDVI_landsat7',3:'NDBI_landsat7',4:'MNDWI_landsat7',\n# 5:'SAVI_landsat7',6:'LSE_landsat7',7:'rad_VIIRS'},inplace=True)\n# dataset_bands['U_class']=dataset_bands.BU_class.apply(lambda y: 1 if y>=3 else 0)\n# dataset_bands['LSE_der_landsat7']=dataset_bands.NDVI_landsat7.apply(lambda y: 0.995 if y < -0.185 else (\n# 0.970 if y< 0.157 else (1.0098+0.047*np.log(y) if y<0.727 else 0.990)))\n# dataset=dataset_bands.reset_index()[['rad_VIIRS','SAVI_landsat7','MNDWI_landsat7','NDBI_landsat7','NDVI_landsat7','LSE_der_landsat7']]\n# class_2014_=rfclassifier.predict(dataset.values)\n# temp_ = rasterio.open('/content/gdrive/My Drive/Urban_monitoring/Urban_/compiled_GHSL_30_train_2014_'+lokasi+'.tif')\n# temp = temp_.read(1)\n# array_class_2014=class_2014_.reshape(temp.shape[0],temp.shape[1]).astype(np.uint8)\n# profile = temp_.profile\n# profile.update(\n# dtype=array_class_2014.dtype,\n# count=1,\n# compress='lzw')\n# with rasterio.open(\n# '/content/gdrive/My Drive/Urban_monitoring/Urban_/GHSL_/GHSL_rev/rev_class_2014_'+lokasi+'.tif','w',**profile) as dst2:\n# dst2.write(array_class_2014,1)\n# dst2.close()", "bandungraya\n" ], [ "confussion_matrix_list[5]", "_____no_output_____" ] ] ]

[ "markdown", "code" ]

[ [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "'''\n\nManual AppleMusic Scraper\n\n'''", "_____no_output_____" ], [ "from selenium import webdriver\nfrom selenium.webdriver.common.keys import Keys\nimport time\nimport datetime\nimport json\n\ndriver = webdriver.Chrome()\ndriver.get(\"https://music.apple.com/\")", "_____no_output_____" ] ], [ [ "### Sign in to Apple Music and navigate to required song class-name", "_____no_output_____" ] ], [ [ "songs_wrap = driver.find_elements_by_class_name(\"song-name\")\nsongs = [x.text for x in song_wrap]\n\nsong_subdiv = driver.find_elements_by_class_name(\"song-name-wrapper\")\n\nartists_wrap = [x.find_element_by_class_name(\"dt-link-to\") for x in song_subdiv]\nartists = [x.text for x in artists_wrap]", "_____no_output_____" ], [ "len(songs)", "_____no_output_____" ], [ "len(artists)", "_____no_output_____" ], [ "'''\nTo combine song and artists lists into a single list of dictionaries\n'''\nl = []\nd = {}\n\nfor x in range(0,99):\n d[\"artist\"] = artists[x]\n d[\"song\"] = songs[x]\n l.append(d)\n d = {} ", "_____no_output_____" ], [ "l", "_____no_output_____" ], [ "#Write list as Json file\nwith open(\"Love100++.json\", \"w\") as read_file:\n json.dump(l, read_file)", "_____no_output_____" ] ] ]

[ "code", "markdown", "code" ]

[ [ "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "Copyright 2021 Google LLC\n\nLicensed under the Apache License, Version 2.0 (the \"License\");\nyou may not use this file except in compliance with the License.\nYou may obtain a copy of the License at\n\n https://www.apache.org/licenses/LICENSE-2.0\n\nUnless required by applicable law or agreed to in writing, software\ndistributed under the License is distributed on an \"AS IS\" BASIS,\nWITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\nSee the License for the specific language governing permissions and\nlimitations under the License.", "_____no_output_____" ], [ "# Quantum Neural Networks\n\nThis notebook provides an introduction to Quantum Neural Networks (QNNs) using the Cirq. The presentation mostly follows [Farhi and Neven](https://arxiv.org/abs/1802.06002). We will construct a simple network for classification to demonstrate its utility on some randomly generated toy data.\n\nFirst we need to install cirq, which has to be done each time this notebook is run. Executing the following cell will do that.", "_____no_output_____" ] ], [ [ "# install published dev version\n# !pip install cirq~=0.4.0.dev\n\n# install directly from HEAD:\n!pip install git+https://github.com/quantumlib/Cirq.git@8c59dd97f8880ac5a70c39affa64d5024a2364d0", "Collecting git+https://github.com/quantumlib/Cirq.git@8c59dd97f8880ac5a70c39affa64d5024a2364d0\n Cloning https://github.com/quantumlib/Cirq.git (to revision 8c59dd97f8880ac5a70c39affa64d5024a2364d0) to /tmp/pip-req-build-p85k13_x\nRequirement already satisfied: google-api-python-client~=1.6 in /usr/local/lib/python3.6/dist-packages (from cirq==0.4.0.dev42) (1.6.7)\nCollecting matplotlib~=2.2 (from cirq==0.4.0.dev42)\n\u001b[?25l Downloading https://files.pythonhosted.org/packages/9e/59/f235ab21bbe7b7c6570c4abf17ffb893071f4fa3b9cf557b09b60359ad9a/matplotlib-2.2.3-cp36-cp36m-manylinux1_x86_64.whl (12.6MB)\n\u001b[K 100% |████████████████████████████████| 12.6MB 837kB/s \n\u001b[?25hRequirement already satisfied: networkx~=2.1 in /usr/local/lib/python3.6/dist-packages (from cirq==0.4.0.dev42) (2.2)\nRequirement already satisfied: numpy~=1.12 in /usr/local/lib/python3.6/dist-packages (from cirq==0.4.0.dev42) (1.14.6)\nRequirement already satisfied: protobuf~=3.5 in /usr/local/lib/python3.6/dist-packages (from cirq==0.4.0.dev42) (3.6.1)\nRequirement already satisfied: requests~=2.18 in /usr/local/lib/python3.6/dist-packages (from cirq==0.4.0.dev42) (2.18.4)\nCollecting sortedcontainers~=2.0 (from cirq==0.4.0.dev42)\n Downloading https://files.pythonhosted.org/packages/be/e3/a065de5fdd5849450a8a16a52a96c8db5f498f245e7eda06cc6725d04b80/sortedcontainers-2.0.5-py2.py3-none-any.whl\nRequirement already satisfied: scipy in /usr/local/lib/python3.6/dist-packages (from cirq==0.4.0.dev42) (1.1.0)\nCollecting typing_extensions (from cirq==0.4.0.dev42)\n Downloading https://files.pythonhosted.org/packages/62/4f/392a1fa2873e646f5990eb6f956e662d8a235ab474450c72487745f67276/typing_extensions-3.6.6-py3-none-any.whl\nRequirement already satisfied: oauth2client<5.0.0dev,>=1.5.0 in /usr/local/lib/python3.6/dist-packages (from google-api-python-client~=1.6->cirq==0.4.0.dev42) (4.1.3)\nRequirement already satisfied: uritemplate<4dev,>=3.0.0 in /usr/local/lib/python3.6/dist-packages (from google-api-python-client~=1.6->cirq==0.4.0.dev42) (3.0.0)\nRequirement already satisfied: six<2dev,>=1.6.1 in /usr/local/lib/python3.6/dist-packages (from google-api-python-client~=1.6->cirq==0.4.0.dev42) (1.11.0)\nRequirement already satisfied: httplib2<1dev,>=0.9.2 in /usr/local/lib/python3.6/dist-packages (from google-api-python-client~=1.6->cirq==0.4.0.dev42) (0.11.3)\nRequirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /usr/local/lib/python3.6/dist-packages (from matplotlib~=2.2->cirq==0.4.0.dev42) (2.3.0)\nRequirement already satisfied: python-dateutil>=2.1 in /usr/local/lib/python3.6/dist-packages (from matplotlib~=2.2->cirq==0.4.0.dev42) (2.5.3)\nRequirement already satisfied: pytz in /usr/local/lib/python3.6/dist-packages (from matplotlib~=2.2->cirq==0.4.0.dev42) (2018.7)\nRequirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.6/dist-packages (from matplotlib~=2.2->cirq==0.4.0.dev42) (0.10.0)\nCollecting kiwisolver>=1.0.1 (from matplotlib~=2.2->cirq==0.4.0.dev42)\n\u001b[?25l Downloading https://files.pythonhosted.org/packages/69/a7/88719d132b18300b4369fbffa741841cfd36d1e637e1990f27929945b538/kiwisolver-1.0.1-cp36-cp36m-manylinux1_x86_64.whl (949kB)\n\u001b[K 100% |████████████████████████████████| 952kB 10.0MB/s \n\u001b[?25hRequirement already satisfied: decorator>=4.3.0 in /usr/local/lib/python3.6/dist-packages (from networkx~=2.1->cirq==0.4.0.dev42) (4.3.0)\nRequirement already satisfied: setuptools in /usr/local/lib/python3.6/dist-packages (from protobuf~=3.5->cirq==0.4.0.dev42) (40.6.2)\nRequirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.6/dist-packages (from requests~=2.18->cirq==0.4.0.dev42) (2018.10.15)\nRequirement already satisfied: idna<2.7,>=2.5 in /usr/local/lib/python3.6/dist-packages (from requests~=2.18->cirq==0.4.0.dev42) (2.6)\nRequirement already satisfied: chardet<3.1.0,>=3.0.2 in /usr/local/lib/python3.6/dist-packages (from requests~=2.18->cirq==0.4.0.dev42) (3.0.4)\nRequirement already satisfied: urllib3<1.23,>=1.21.1 in /usr/local/lib/python3.6/dist-packages (from requests~=2.18->cirq==0.4.0.dev42) (1.22)\nRequirement already satisfied: pyasn1>=0.1.7 in /usr/local/lib/python3.6/dist-packages (from oauth2client<5.0.0dev,>=1.5.0->google-api-python-client~=1.6->cirq==0.4.0.dev42) (0.4.4)\nRequirement already satisfied: rsa>=3.1.4 in /usr/local/lib/python3.6/dist-packages (from oauth2client<5.0.0dev,>=1.5.0->google-api-python-client~=1.6->cirq==0.4.0.dev42) (4.0)\nRequirement already satisfied: pyasn1-modules>=0.0.5 in /usr/local/lib/python3.6/dist-packages (from oauth2client<5.0.0dev,>=1.5.0->google-api-python-client~=1.6->cirq==0.4.0.dev42) (0.2.2)\nBuilding wheels for collected packages: cirq\n Running setup.py bdist_wheel for cirq ... \u001b[?25l-\b \b\\\b \b|\b \b/\b \bdone\n\u001b[?25h Stored in directory: /tmp/pip-ephem-wheel-cache-ub0vqjao/wheels/ae/db/83/edbb28e59157c931c9342e7ae581bebdf4939c0465a413aaaf\nSuccessfully built cirq\nInstalling collected packages: kiwisolver, matplotlib, sortedcontainers, typing-extensions, cirq\n Found existing installation: matplotlib 2.1.2\n Uninstalling matplotlib-2.1.2:\n Successfully uninstalled matplotlib-2.1.2\nSuccessfully installed cirq-0.4.0.dev42 kiwisolver-1.0.1 matplotlib-2.2.3 sortedcontainers-2.0.5 typing-extensions-3.6.6\n" ] ], [ [ "To verify that Cirq is installed in your environment, try to `import cirq` and print out a diagram of the Foxtail device. It should produce a 2x11 grid of qubits.", "_____no_output_____" ] ], [ [ "import cirq\nimport numpy as np\nimport matplotlib.pyplot as plt\n\nprint(cirq.google.Foxtail)", "(0, 0)───(0, 1)───(0, 2)───(0, 3)───(0, 4)───(0, 5)───(0, 6)───(0, 7)───(0, 8)───(0, 9)───(0, 10)\n│ │ │ │ │ │ │ │ │ │ │\n│ │ │ │ │ │ │ │ │ │ │\n(1, 0)───(1, 1)───(1, 2)───(1, 3)───(1, 4)───(1, 5)───(1, 6)───(1, 7)───(1, 8)───(1, 9)───(1, 10)\n" ] ], [ [ "### The QNN Idea\nWe'll begin by describing here the QNN model we are pursuing. We'll discuss the quantum circuit describing a very simple neuron, and how it can be trained.", "_____no_output_____" ], [ "As in an ordinary neural network, a QNN takes in data, processes that data, and then returns an answer. In the quantum case, the data will be encoded into the initial quantum state, and the processing step is the action of a quantum circuit on that quantum state. At the end we will measure one or more of the qubits, and the statistics of those measurements are the output of the net.\n\n", "_____no_output_____" ], [ "#### Classical vs Quantum", "_____no_output_____" ], [ "An ordinary neural network can only handle classical input. The input to a QNN, though, is a quantum state, which consists of $2^n$ complex amplitudes for $n$-qubits. If you attached your quantum computer directly to some physics experiment, for example, then you could have a QNN do some post-processing on the experimental wavefunction in lieu of a more traditional measurement. There are some very exciting possiblities there, but unfortunately we wil not be considering them in this Colab. It requires significantly more quantum background to understand what's going on, and it's harder to give examples because the input states themselves can be quite complicated. For recent examples of that kind of network, though, check out [this](https://arxiv.org/abs/1805.08654) paper and [this](https://arxiv.org/abs/1810.03787) paper. The basic ingredients are similar to what we'll cover here", "_____no_output_____" ], [ "In this Colab we'll focus on classical inputs, by which I mean the specification of one of the computational basis states as the initial state. There are a total of $2^n$ of these states for $n$ qubits. Note the crucial difference between this case and the quantum case: in the quantum case the input is $2^n$-*dimensional*, while in the classical case there are $2^n$ *possible* inputs. The quantum neural network can process these inputs in a \"quantum\" way, meaning that it may be able to evaluate certain functions on these inputs more efficiently than a classical network. Whether the \"quantum\" processing is actually useful in practice remains to be seen, and in this Colab we will not have time to really get into that aspect of a QNN.", "_____no_output_____" ], [ "#### Data Processing", "_____no_output_____" ], [ "Given the classical input state, what will we do with it? At this stage it's helpful to be more specific and definite about the problem we are trying to solve. The problem we're going to focus on in this Colab is __two-category classicfication__. That means that after the quantum circuit has finished running, we measure one of the qubits, the *readout* qubit, and the value of that qubit will tell us which of the two categories our classical input state belonged to. Since this is quantum, the output that qubit is going to be random according to some probability distributuion. So really we're going to repeat the computation many times and take a majority vote.", "_____no_output_____" ], [ "Our classical input data is a bitstring that is converted into a computational basis state. We want to influence the readout qubit in a way that depends on this state. Our main tool for this a gate we call the $ZX$-gate, which acts on two qubits as\n$$\n\\exp(i \\pi w Z \\otimes X) = \\begin{bmatrix}\n\\cos \\pi w & i\\sin\\pi w &0&0\\\\\ni\\sin\\pi w & \\cos \\pi w &0&0\\\\\n0&0& \\cos \\pi w & -i\\sin\\pi w \\\\\n0&0 & -i\\sin\\pi w & \\cos\\pi w \n\\end{bmatrix},\n$$\nwhere $w$ is a free parameter ($w$ stands for weight). This gate rotates the second qubit around the $X$-axis (on the Bloch sphere) either clockwise or counterclockwise depending on the state of the first qubit as seen in the computational basis. The amount of the rotation is determined by $w$.", "_____no_output_____" ], [ "If we connect each of our input qubits to the readout qubit using one of these gates, then the result is that the readout qubit will be rotated in a way that depeonds the initial state in a straightforward way. This rotation is in the $YZ$ plane, so will change the statistics of measurements in either the $Z$ basis or the $Y$ basis for the readout qubit. We're going to choose to have the initial state of the readout qubit to be a standard computational basis state as usual, which is a $Z$ eigenstate but \"neutral\" with respect to $Y$ (i.e., 50/50 probabilty of $Y=+1$ or $Y=-1$). Then after all of the rotations are complete we'll measure the readout qubit in the $Y$ basis. If all goes well, then the net rotation induced by the $ZX$ gates will place the readout qubit near one of the two $Y$ eigenstates in a way that depends on the initial data.", "_____no_output_____" ], [ "To summarize, here is our strategy for processing the two-category classification problem:\n\n1) Prepare a computational basis state corresponding to the input that should be categorized.\n\n2) Use $ZX$ gates to rotate the state of the readout qubit in a way that depends on the input.\n\n3) Measure the readout qubit in the $Y$ basis to get the predicted label. Take a majority vote after many repetitions.", "_____no_output_____" ], [ "This is the simplest possible kind of network, and really only corresponds to a single neuron. We'll talk about more complicated possibilities after understanding how to implement this one.", "_____no_output_____" ], [ "### Custom Two-Qubit Gate\n\nOur first task is to code up the $ZX$ gate described above, which is given by the matrix\n$$\n\\exp(i \\pi w Z \\otimes X) = \\begin{bmatrix}\n\\cos \\pi w & i\\sin\\pi w &0&0\\\\\ni\\sin\\pi w & \\cos \\pi w &0&0\\\\\n0&0& \\cos \\pi w & -i\\sin\\pi w \\\\\n0&0 & -i\\sin\\pi w & \\cos\\pi w \n\\end{bmatrix},\n$$", "_____no_output_____" ], [ "Just from the form of the gate we can see that it performs a rotation by angle $\\pm \\pi w$ on the second qubit depending on the value of the first qubit. If we only had one or the other of these two blocks, then this gate would literally be a controlled rotation. For example, using the Cirq conventions,\n$$\nCR_X(\\theta) = \\begin{bmatrix}\n1 & 0 &0&0\\\\\n0 & 1 &0&0\\\\\n0&0& \\cos \\theta/2 & -i\\sin \\theta/2 \\\\\n0&0 & -i\\sin\\theta/2 & \\cos\\theta/2 \n\\end{bmatrix},\n$$\nwhich means that setting $\\theta = 2\\pi w$ should give us (part) of our desired transformation.", "_____no_output_____" ] ], [ [ "a = cirq.NamedQubit(\"a\")\nb = cirq.NamedQubit(\"b\")\nw = .25 # Put your own weight here.\nangle = 2*np.pi*w\ncircuit = cirq.Circuit.from_ops(cirq.ControlledGate(cirq.Rx(angle)).on(a,b))\nprint(circuit)\ncircuit.to_unitary_matrix().round(2)", "a: ───@──────────\n │\nb: ───Rx(0.5π)───\n" ] ], [ [ "__Question__: The rotation in the upper-left block is by the opposite angle. But how do we get the rotation to happen in the upper-left block of the $4\\times 4$ matrix in the first place? What is the circuit?", "_____no_output_____" ], [ "#### Solution", "_____no_output_____" ], [ "Switching the upper-left and lower-right blocks of a controlled gate corresponds to activating when the control qubit is in the $|0\\rangle$ state instead of the $|1\\rangle$ state. We can arrange this to happen by taking the control gate we already have and conjugating the control qubit by $X$ gates (which implement the NOT operation). Don't forget to also rotate by the opposite angle.", "_____no_output_____" ] ], [ [ "a = cirq.NamedQubit(\"a\")\nb = cirq.NamedQubit(\"b\")\nw = 0.25 # Put your own weight here.\nangle = 2*np.pi*w\ncircuit = cirq.Circuit.from_ops([cirq.X(a),\n cirq.ControlledGate(cirq.Rx(-angle)).on(a,b),\n cirq.X(a)])\nprint(circuit)\ncircuit.to_unitary_matrix().round(2)", "a: ───X───@───────────X───\n │\nb: ───────Rx(-0.5π)───────\n" ] ], [ [ "#### The Full $ZX$ Gate", "_____no_output_____" ], [ "We can put together the two controlled rotations to make the full $ZX$ gate. Having discussed the decomposition already, we can make our own class and specify its action using the `_decpompose_` method. Fill in the following code block to implement this gate.", "_____no_output_____" ] ], [ [ "class ZXGate(cirq.ops.gate_features.TwoQubitGate):\n \"\"\"ZXGate with variable weight.\"\"\"\n \n def __init__(self, weight=1):\n \"\"\"Initializes the ZX Gate up to phase.\n\n Args:\n weight: rotation angle, period 2 \n \"\"\"\n self.weight = weight\n \n def _decompose_(self, qubits):\n a, b = qubits\n ## YOUR CODE HERE\n\n # This lets the weight be a Symbol. Useful for paramterization.\n def _resolve_parameters_(self, param_resolver):\n return ZXGate(weight=param_resolver.value_of(self.weight))\n\n # How should the gate look in ASCII diagrams?\n def _circuit_diagram_info_(self, args): \n return cirq.protocols.CircuitDiagramInfo(\n wire_symbols=('Z', 'X'),\n exponent=self.weight) ", "_____no_output_____" ] ], [ [ "#### Solution", "_____no_output_____" ] ], [ [ "class ZXGate(cirq.ops.gate_features.TwoQubitGate):\n \"\"\"ZXGate with variable weight.\"\"\"\n \n def __init__(self, weight=1):\n \"\"\"Initializes the ZX Gate up to phase.\n\n Args:\n weight: rotation angle, period 2 \n \"\"\"\n self.weight = weight\n \n def _decompose_(self, qubits):\n a, b = qubits\n yield cirq.ControlledGate(cirq.Rx(2*np.pi*self.weight)).on(a,b)\n yield cirq.X(a)\n yield cirq.ControlledGate(cirq.Rx(-2*np.pi*self.weight)).on(a,b)\n yield cirq.X(a)\n\n # This lets the weight be a Symbol. Useful for paramterization.\n def _resolve_parameters_(self, param_resolver):\n return ZXGate(weight=param_resolver.value_of(self.weight))\n\n # How should the gate look in ASCII diagrams?\n def _circuit_diagram_info_(self, args): \n return cirq.protocols.CircuitDiagramInfo(\n wire_symbols=('Z', 'X'),\n exponent=self.weight) ", "_____no_output_____" ] ], [ [ "#### EigenGate Implementation\n\nAnother way to specify how a gate works is by an explicit eigen-action. In our case that is also easy, since we know that the gate acts as a phase (the eigenvalue) when the first qubit is in a $Z$ eigenstate (i.e., a computational basis state) and the second qubit is in an $X$ eigenstate.\n\nThe way we specify eigen-actions in Cirq is through the `_eigen_components` method, where we need to specify the eigenvalue as a phase together with a projector onto the eigenspace of that phase. We also conventionally specify the gate at $w=1$ and set $w$ internally to be the `exponent` of the gate, which automatically implements other values of $w$ for us.\n\nIn the case of the $ZX$ gate with $w=1$, one of our eigenvalues is $\\exp(+i\\pi)$, which is specified as $1$ in Cirq. (Because $1$ is the coefficeint of $i\\pi$ in the exponential.) This is the phase when when the first qubit is in the $Z=+1$ state and the second qubit is in the $X=+1$ state, or when the first qubit is in the $Z=-1$ state and the second qubit is in the $X=-1$ state. The projector onto these states is\n$$\n\\begin{align}\nP &= |0+\\rangle \\langle 0{+}| + |1-\\rangle \\langle 1{-}|\\\\\n&= \\frac{1}{2}\\big(|00\\rangle \\langle 00| +|00\\rangle \\langle 01|+|01\\rangle \\langle 00|+|01\\rangle \\langle 01|+ |10\\rangle \\langle 10|-|10\\rangle \\langle 11|-|11\\rangle \\langle 10|+|11\\rangle \\langle 11|\\big)\\\\\n&=\\frac{1}{2}\\begin{bmatrix}\n1 & 1 &0&0\\\\\n1 & 1 &0&0\\\\\n0&0& 1 & -1 \\\\\n0&0 & -1 & 1\n\\end{bmatrix}\n\\end{align}\n$$\nA similar formula holds for the eigenvalue $\\exp(-i\\pi)$ with the two blocks in the projector flipped.\n\n\n__Exercise__: Implement the $ZX$ gate as an `EigenGate` using this decomposition. The following codeblock will get you started.\n", "_____no_output_____" ] ], [ [ "class ZXGate(cirq.ops.eigen_gate.EigenGate,\n cirq.ops.gate_features.TwoQubitGate):\n \"\"\"ZXGate with variable weight.\"\"\"\n \n def __init__(self, weight=1):\n \"\"\"Initializes the ZX Gate up to phase.\n\n Args:\n weight: rotation angle, period 2 \n \"\"\"\n self.weight = weight\n super().__init__(exponent=weight) # Automatically handles weights other than 1\n \n def _eigen_components(self):\n return [\n (1, np.array([[0.5, 0.5, 0, 0],\n [ 0.5, 0.5, 0, 0],\n [0, 0, 0.5, -0.5],\n [0, 0, -0.5, 0.5]])),\n (??, ??) # YOUR CODE HERE: phase and projector for the other eigenvalue\n ]\n\n # This lets the weight be a Symbol. Useful for parameterization.\n def _resolve_parameters_(self, param_resolver):\n return ZXGate(weight=param_resolver.value_of(self.weight))\n\n # How should the gate look in ASCII diagrams?\n def _circuit_diagram_info_(self, args): \n return cirq.protocols.CircuitDiagramInfo(\n wire_symbols=('Z', 'X'),\n exponent=self.weight) ", "_____no_output_____" ] ], [ [ "#### Solution", "_____no_output_____" ] ], [ [ "class ZXGate(cirq.ops.eigen_gate.EigenGate,\n cirq.ops.gate_features.TwoQubitGate):\n \"\"\"ZXGate with variable weight.\"\"\"\n \n def __init__(self, weight=1):\n \"\"\"Initializes the ZX Gate up to phase.\n\n Args:\n weight: rotation angle, period 2 \n \"\"\"\n self.weight = weight\n super().__init__(exponent=weight) # Automatically handles weights other than 1\n \n def _eigen_components(self):\n return [\n (1, np.array([[0.5, 0.5, 0, 0],\n [ 0.5, 0.5, 0, 0],\n [0, 0, 0.5, -0.5],\n [0, 0, -0.5, 0.5]])),\n (-1, np.array([[0.5, -0.5, 0, 0],\n [ -0.5, 0.5, 0, 0],\n [0, 0, 0.5, 0.5],\n [0, 0, 0.5, 0.5]]))\n ]\n\n # This lets the weight be a Symbol. Useful for parameterization.\n def _resolve_parameters_(self, param_resolver):\n return ZXGate(weight=param_resolver.value_of(self.weight))\n\n # How should the gate look in ASCII diagrams?\n def _circuit_diagram_info_(self, args): \n return cirq.protocols.CircuitDiagramInfo(\n wire_symbols=('Z', 'X'),\n exponent=self.weight) ", "_____no_output_____" ] ], [ [ "#### Testing the Gate", "_____no_output_____" ], [ "__BEFORE MOVING ON__ make sure you've executed the `EigenGate` solution of the $ZX$ gate implementation. That's the one assumed for the code below, though other implementations may work just as well. In general, the cells in this Colab may depend on previous cells.\n\nLet's test out our gate. First we'll make a simple test circuit to see that the ASCII diagrams are diplaying properly:", "_____no_output_____" ] ], [ [ "a = cirq.NamedQubit(\"a\")\nb = cirq.NamedQubit(\"b\")\nw = .15 # Put your own weight here. Try using a cirq.Symbol.\ncircuit = cirq.Circuit.from_ops(ZXGate(w).on(a,b))\nprint(circuit)", "a: ───Z────────\n │\nb: ───X^0.15───\n" ] ], [ [ "We should also check that the matrix is what we expect:", "_____no_output_____" ] ], [ [ "test_matrix = np.array([[np.cos(np.pi*w), 1j*np.sin(np.pi*w), 0, 0],\n [1j*np.sin(np.pi*w), np.cos(np.pi*w), 0, 0],\n [0, 0, np.cos(np.pi*w), -1j*np.sin(np.pi*w)],\n [0, 0, -1j*np.sin(np.pi*w),np.cos(np.pi*w)]])\n# Test for five digits of accuracy. Won't work with cirq.Symbol\nassert (circuit.to_unitary_matrix().round(5) == test_matrix.round(5)).all()", "_____no_output_____" ] ], [ [ "### Create Circuit\n\nNow we have to create the QNN circuit. We are simply going to let a $ZX$ gate act between each data qubit and the readout qubit. For simplicity, let's share a single weight between all of the gates. You are invited to experiment with making the weights different, but in our example problem below we can set them all equal by symmetry.\n\n__Question__: What about the order of these actions? Which data qubits should interact with the readout qubit first?\n\nRemember that we also want to measure the readout qubit in the $Y$ basis. Fundamentally speaking, all measurements in Cirq are computational basis measurements, and so we have to implement the change of basis by hand.\n\n\n__Question__: What is the circuit for a basis transformation from the $Y$ basis to the computational basis? We want to choose our transformation so that an eigenstate with $Y=+1$ becomes an eigenstate with $Z=+1$ prior to measurement.", "_____no_output_____" ], [ "#### Solutions", "_____no_output_____" ], [ "* The $ZX$ gates all commute with each other, so the order of implementation doesn't matter!\n\n* We want a transformation that maps $\\big(|0\\rangle + i |1\\rangle\\big)/\\sqrt{2}$ to $|0\\rangle$ and $\\big(|0\\rangle - i |1\\rangle\\big)\\sqrt{2}$ to $|1\\rangle$. Recall that the phase gate $S$ is given in matrix form by\n$$\nS = \\begin{bmatrix}\n1 & 0 \\\\\n0 & i\n\\end{bmatrix},\n$$\nand the Hadamard transform is given by\n$$\nH = \\frac{1}{\\sqrt{2}}\\begin{bmatrix}\n1 & 1 \\\\\n1 & -1\n\\end{bmatrix},\n$$\nSo acting with $S^{-1}$ and then $H$ gives what we want. We'll add these two gates to the end of the circuit on the readout qubit so that the final measurement effectively occurs in the $Y$ basis.\n\n", "_____no_output_____" ], [ "#### Make Circuit", "_____no_output_____" ], [ "A clean way of making circuits is to define generators for logically-related circuit elements, and then `append` those to the circuit you want to make. Here is a code snippet that initializes our qubits and defines a generator for a single layer of $ZX$ gates:", "_____no_output_____" ] ], [ [ "# Total number of data qubits\nINPUT_SIZE = 9\n\ndata_qubits = cirq.LineQubit.range(INPUT_SIZE)\nreadout = cirq.NamedQubit('r')\n\n# Initialize parameters of the circuit\nparams = {'w': 0}\n\ndef ZX_layer():\n \"\"\"Adds a ZX gate between each data qubit and the readout.\n All gates are given the same cirq.Symbol for a weight.\"\"\"\n for qubit in data_qubits:\n yield ZXGate(cirq.Symbol('w')).on(qubit, readout)", "_____no_output_____" ] ], [ [ "Use this generator to create the QNN circuit. Don't forget to add the basis change for the readout qubit at the end!", "_____no_output_____" ] ], [ [ "qnn = cirq.Circuit()\nqnn.append(???) # YOUR CODE HERE", "_____no_output_____" ] ], [ [ "#### Solution ", "_____no_output_____" ] ], [ [ "qnn = cirq.Circuit()\nqnn.append(ZX_layer())\nqnn.append([cirq.S(readout)**-1, cirq.H(readout)]) # Basis transformation", "_____no_output_____" ] ], [ [ "#### View the Circuit", "_____no_output_____" ], [ "It's usually a good idea to view the ASCII diagram of your circuit to make sure it's doing what you want. This can be displayed by printing the circuit.", "_____no_output_____" ] ], [ [ "print(qnn)", "0: ───Z────────────────────────────────────────────────────────────────\n │\n1: ───┼─────Z──────────────────────────────────────────────────────────\n │ │\n2: ───┼─────┼─────Z────────────────────────────────────────────────────\n │ │ │\n3: ───┼─────┼─────┼─────Z──────────────────────────────────────────────\n │ │ │ │\n4: ───┼─────┼─────┼─────┼─────Z────────────────────────────────────────\n │ │ │ │ │\n5: ───┼─────┼─────┼─────┼─────┼─────Z──────────────────────────────────\n │ │ │ │ │ │\n6: ───┼─────┼─────┼─────┼─────┼─────┼─────Z────────────────────────────\n │ │ │ │ │ │ │\n7: ───┼─────┼─────┼─────┼─────┼─────┼─────┼─────Z──────────────────────\n │ │ │ │ │ │ │ │\n8: ───┼─────┼─────┼─────┼─────┼─────┼─────┼─────┼─────Z────────────────\n │ │ │ │ │ │ │ │ │\nr: ───X^w───X^w───X^w───X^w───X^w───X^w───X^w───X^w───X^w───S^-1───H───\n" ] ], [ [ "You can experiment with adding more layers of $ZX$ gates (or adding other kinds of transformations!) to your QNN, but we can use this simplest kind of circuit to analyze a simple toy problem, which is what we will do next.", "_____no_output_____" ], [ "### A Toy Problem: Biased Coin Flips\n\nAs a toy problem, let's get our quantum neuron to decide whether a coin is biased toward heads or toward tails based on a sequence of coin flips.", "_____no_output_____" ], [ "To be specific, let's try to train a QNN to distinguish between a coin that yields \"heads\" with probability $p$, and one that yields \"heads\" with probability $1-p$. Without loss of generality, let's say that $p\\leq 0.5$. We don't need a fancy QNN to come up with a winning strategy: given a series of coin flips, you guess $p$ if the majority of flips are \"tails\" and $1-p$ if the majority are \"heads.\" But for purposes of illustration, let's do it the fancy way.", "_____no_output_____" ], [ "To translate this problem into our QNN language, we need to encode the sequence of coin flips into a computational basis state. Let's associate $0$ with tails and $1$ with heads. So the sequence of coin flips becomes a sequence of $0$s and $1$s, and these define a computational basis state.", "_____no_output_____" ], [ "We also need to define a convention for our labeling of the two coins. We'll say that the $p$ coin (majority tails) gets the label $-1$ and the $1-p$ coin (majority heads) gets the label $+1$. So when we measure $Y$ at the end of the computation we can say that the majority-vote of the $Y$ outcome is our predicted label.", "_____no_output_____" ], [ "To be a little more nuanced (and to aid the formulation of the problem), let's say that the expectation value $\\langle Y \\rangle$ for a given input state defines our estimator for the label of that state. We're going to use that to define a loss function for training next.\n", "_____no_output_____" ], [ "### Define Loss Function", "_____no_output_____" ], [ "Suppose we have a collection of $N$ (bitstring, label) pairs. A useful loss function to characterize the effectiveness of our QNN on this collection is\n$$\n\\text{Loss} = \\frac{1}{2N}\\sum_{j=1}^n (1- \\ell_j\\langle Y \\rangle_j),\n$$\nwhere $\\ell_j$ is the label of the $j$th pair and $\\langle Y \\rangle_j$ is the expectation value of $Y$ on the readout qubit using the $j$th bitstring as input. If the network is perfect, the loss is equal to zero. If the network is maximally unsure about the labels (so that $\\langle Y \\rangle_j = 0$ for all $j$) then the loss is equal to $1/2$. And if the network gets everything wrong, then the loss is equal to $1$. We're going to train our network using this loss function, so next we'll write some functions to compute the loss.", "_____no_output_____" ], [ "Another useful function to have around is the average classification error. Recall that our prescription was to execute the quantum circuit many times and take a majority vote to compute the predicted label. The majority vote for the readout is the same as $\\text{sign}(\\langle Y \\rangle)$, so we can write a formula for the error in this procedure as\n$$\n\\text{Error} = \\frac{1}{2N}\\sum_{j=1}^n \\big(1- \\ell_j\\text{sign}\\big(\\langle Y \\rangle_j\\big)\\big).\n$$\nThis is not so useful as a loss function because it is not smooth and does not provide an incentive to make $|\\langle Y \\rangle|$ large, but it can be an informative quantity to compute.\n\n__Question__: Why would we want $|\\langle Y \\rangle|$ to be large?", "_____no_output_____" ], [ "#### Solution", "_____no_output_____" ], [ "When we implement this algorithm on the actual hardware, $\\langle Y \\rangle$ can only be estimated by repeatedly executing the circuit and measuring the result. The more measurements we make, the better our estimate of $\\langle Y \\rangle$ will be. Even if we are only interested in $\\text{sign}\\big(\\langle Y \\rangle\\big)$, we will need to meake enough measurements to be sure that our estimate has the correct sign, and if $|\\langle Y \\rangle|$ is large then fewer measurements will be required to have high confidence in the sign.\n\n\nFurthermore, if the machine is noisy (which it will be), then the noise will induce some errors in our estimate of $\\langle Y \\rangle$. If $|\\langle Y \\rangle|$ is small then it's likely that the noise will lead to the wrong sign.", "_____no_output_____" ], [ "#### Expectation Value\n\nOur first function computes the expectation value of the readout qubit for our circuit given a specification of the initial state. Rather than a bitstring, we'll specify the initial state as an array of $0$s and $1$s. These are the outputs of the coin flips in our toy problem. We'll compute the expectation value exactly using the wavefunction for now.", "_____no_output_____" ] ], [ [ "def readout_expectation(state):\n \"\"\"Takes in a specification of a state as an array of 0s and 1s\n and returns the expectation value of Z on ther readout qubit.\n Uses the XmonSimulator to calculate the wavefunction exactly.\"\"\" \n \n # A convenient representation of the state as an integer\n state_num = int(np.sum(state*2**np.arange(len(state))))\n\n resolver = cirq.ParamResolver(params)\n simulator = cirq.Simulator()\n\n # Specify an explicit qubit order so that we know which qubit is the readout\n result = simulator.simulate(qnn, resolver, qubit_order=[readout]+data_qubits,\n initial_state=state_num)\n wf = result.final_state\n\n # Becase we specified qubit order, the Z value of the readout is the most\n # significant bit.\n Z_readout = np.append(np.ones(2**INPUT_SIZE), -np.ones(2**INPUT_SIZE))\n\n return np.sum(np.abs(wf)**2 * Z_readout)", "_____no_output_____" ] ], [ [ "#### Loss and Error\n\nThe next functions take a list of states (each specified as an array of $0$s and $1$s as before) and a corresponding list of labels and computes the loss and error, respectively, of that list.", "_____no_output_____" ] ], [ [ "def loss(states, labels):\n loss=0\n for state, label in zip(states,labels):\n loss += 1 - label*readout_expectation(state)\n return loss/(2*len(states))\n \ndef classification_error(states, labels):\n error=0\n for state,label in zip(states,labels):\n error += 1 - label*np.sign(readout_expectation(state))\n return error/(2*len(states))", "_____no_output_____" ] ], [ [ "#### Generating Data", "_____no_output_____" ], [ "For our toy problem we'll want to be able to generate a batch of data. Here is a helper function for that task:", "_____no_output_____" ] ], [ [ "def make_batch():\n \"\"\"Generates a set of labels, then uses those labels to generate inputs.\n label = -1 corresponds to majority 0 in the sate, label = +1 corresponds to\n majority 1.\n \"\"\"\n np.random.seed(0) # For consistency in demo\n labels = (-1)**np.random.choice(2, size=100) # Smaller batch sizes will speed up computation\n states = []\n for label in labels:\n states.append(np.random.choice(2, size=INPUT_SIZE, p=[0.5-label*0.2,0.5+label*0.2]))\n return states, labels\n\nstates, labels = make_batch()", "_____no_output_____" ] ], [ [ "### Training\n\nNow we'll try to find the optimal weight to solve our toy problem. For illustration, we'll do both a brute force search of the paramter space as well as a stochastic gradient descent.", "_____no_output_____" ], [ "#### Brute Force Search\n\nLet's compute both the loss and error rate on a batch of data as a function of the shared weight between all the gates.", "_____no_output_____" ] ], [ [ "# Using cirq.Simulator with the EigenGate implementation of ZZ, this takes\n# about 30s to run. Using the XmonSimulator took about 40 minutes the last\n# time I tried it!\n%%time\nlinspace = np.linspace(start=-1, stop=1, num=80)\ntrain_losses = []\nerror_rates = []\nfor p in linspace:\n params = {'w': p}\n train_losses.append(loss(states, labels))\n error_rates.append(classification_error(states, labels))", "CPU times: user 29.5 s, sys: 2.93 ms, total: 29.5 s\nWall time: 29.6 s\n" ], [ "plt.plot(linspace, train_losses)\nplt.xlabel('Weight')\nplt.ylabel('Loss')\nplt.title('Loss as a Function of Weight')\nplt.show()\nplt.plot(linspace, error_rates)\nplt.xlabel('Weight')\nplt.ylabel('Error Rate')\nplt.title('Error Rate as a Function of Weight')\nplt.show()", "_____no_output_____" ] ], [ [ "__Question__: Why are the loss and error functions periodic with period $1$ when the $ZX$ gate is periodic with period $2$?", "_____no_output_____" ], [ "#### Solution", "_____no_output_____" ], [ "This kind of \"halving\" of the periodicity of $\\langle Y \\rangle$ compared to the period of the gates itself is typical of qubit systems. We can analyze how it works mathematically in a simpler setting. Instead of the $ZX$ Gate, let's just imagine that we rotate the readout qubit around the $X$ axis by some fixed amout. This is the effective calculation for a single fixed data input.\n$$\n\\begin{align}\n\\langle Y \\rangle &= \\langle 0 |\\exp(-i \\pi w X) Y \\exp(i \\pi w X) |0 \\rangle\\\\\n&= \\langle 0 |\\big(\\cos \\pi w - i X\\sin \\pi w \\big) Y \\big(\\cos \\pi w + i X \\sin \\pi w \\big) |0 \\rangle\\\\\n&= \\langle 0 |\\big(Y\\cos 2\\pi w +Z \\sin 2\\pi w \\big) |0 \\rangle\\\\\n&= \\sin 2\\pi w.\n\\end{align}\n$$", "_____no_output_____" ], [ "#### Stochastic Gradient Descent\n\nTo train the network we'll use stochastic gradient descent. Note that this isn't necessarily a good idea since the loss function is far from convex, and there's a good chance we'll get stuck in very inefficient local minimum if we initialize the paramters randomly. But as an exercise we'll do it anyway. In the next section we'll discuss other ways to train these sorts of networks.", "_____no_output_____" ], [ "We'll compute the gradient of the loss function using a symmetric finite-difference approximation: $f'(x) \\approx (f(x + \\epsilon) - f(x-\\epsilon))/2\\epsilon$. This is the most straightforward way to do it using the quantum computer. We'll also generate a new instance of the problem each time.", "_____no_output_____" ] ], [ [ "def stochastic_grad_loss():\n \"\"\"Generates a new data point and computes the gradient of the loss\n using that data point.\"\"\"\n \n # Randomly generate the data point.\n label = (-1)**np.random.choice(2)\n state = np.random.choice(2, size=INPUT_SIZE, p=[0.5-label*0.2,0.5+label*0.2])\n \n # Compute the gradient using finite difference\n eps = 10**-5 # Discretization of gradient. Try different values.\n params['w'] -= eps\n loss1 = loss([state],[label])\n params['w'] += 2*eps\n grad = (loss([state],[label])-loss1)/(2*eps)\n params['w'] -= eps # Reset the parameter value\n return grad", "_____no_output_____" ] ], [ [ "We can apply this function repeatedly to flow toward the minimum:", "_____no_output_____" ] ], [ [ "eta = 10**-4 # Learning rate. Try different values.\nparams = {'w': 0} # Initialize weight. Try different values.\nfor i in range(201):\n if not i%25:\n print('Step: {} Loss: {}'.format(i, loss(states, labels)))\n grad = stochastic_grad_loss()\n params['w'] += -eta*grad\nprint('Final Weight: {}'.format(params['w']))", "Step: 0 Loss: 0.5\nStep: 25 Loss: 0.29664170142263174\nStep: 50 Loss: 0.21596111725317313\nStep: 75 Loss: 0.19353972657117993\nStep: 100 Loss: 0.1930989919230342\nStep: 125 Loss: 0.19318223176524044\nStep: 150 Loss: 0.19358215024578385\nStep: 175 Loss: 0.1965144828868506\nStep: 200 Loss: 0.1930640292633325\nFinal Weight: -0.0443500901565141\n" ] ], [ [ "### Use Sampling Instead of Calculating from the Wavefunction\n\nOn real hardware we will have to use sampling to find results instead of computing the exact wavefunction. Rewrite the `readout_expectation` function to compute the expectation value using sampling instead. Unlike with the wavefunction calculation, we also need to build our circuit in a way that accounts for the initial state (we are always assumed to start in the all $|0\\rangle$ state)\n\n\n", "_____no_output_____" ] ], [ [ "def readout_expectation_sample(state):\n \"\"\"Takes in a specification of a state as an array of 0s and 1s\n and returns the expectation value of Z on ther readout qubit.\n Uses the XmonSimulator to sample the final wavefunction.\"\"\"\n \n # We still need to resolve the parameters in the circuit.\n resolver = cirq.ParamResolver(params)\n \n # Make a copy of the QNN to avoid making changes to the global variable.\n measurement_circuit = qnn.copy()\n \n # Modify the measurement circuit to account for the desired input state. \n # YOUR CODE HERE\n \n # Add appropriate measurement gate(s) to the circuit.\n # YOUR CODE HERE\n \n simulator = cirq.google.XmonSimulator() \n result = simulator.run(measurement_circuit, resolver, repetitions=10**6) # Try adjusting the repetitions\n \n # Return the Z expectation value\n return ((-1)**result.measurements['m']).mean()", "_____no_output_____" ] ], [ [ "#### Solution", "_____no_output_____" ] ], [ [ "def readout_expectation_sample(state):\n \"\"\"Takes in a specification of a state as an array of 0s and 1s\n and returns the expectation value of Z on ther readout qubit.\n Uses the XmonSimulator to sample the final wavefunction.\"\"\"\n \n # We still need to resolve the parameters in the circuit.\n resolver = cirq.ParamResolver(params)\n \n # Make a copy of the QNN to avoid making changes to the global variable.\n measurement_circuit = qnn.copy()\n \n # Modify the measurement circuit to account for the desired input state. \n for i, qubit in enumerate(data_qubits):\n if state[i]:\n measurement_circuit.insert(0,cirq.X(qubit))\n \n # Add appropriate measurement gate(s) to the circuit.\n measurement_circuit.append(cirq.measure(readout, key='m'))\n \n simulator = cirq.Simulator() \n result = simulator.run(measurement_circuit, resolver, repetitions=10**6) # Try adjusting the repetitions\n \n # Return the Z expectation value\n return ((-1)**result.measurements['m']).mean()", "_____no_output_____" ] ], [ [ "#### Comparison of Sampling with the Exact Wavefunction", "_____no_output_____" ], [ "Just to illustrate the difference between sampling and using the wavefunction, try running the two methods several times on identical input:\n", "_____no_output_____" ] ], [ [ "state = [0,0,0,1,0,1,1,0,1] # Try different initial states.\nparams = {'w': 0.05} # Try different weights.\n\nprint(\"Exact expectation value: {}\".format(readout_expectation(state)))\nprint(\"Estimates from sampling:\")\nfor _ in range(5):\n print(readout_expectation_sample(state))", "Exact expectation value: 0.3090169429779053\nEstimates from sampling:\n0.308354\n0.306674\n0.309026\n0.310854\n0.30854\n" ] ], [ [ "As an exercise, try repeating some of the above calculations (e.g., the SGD optimization) using `readout_expectation_sample` in place of `readout_expectation`. How many repetitions should you use? How should the hyperparameters `eps` and `eta` be adjusted in response to the number of repetitions?", "_____no_output_____" ], [ "### Optimizing For Hardware\n\nThere are more issues to think about if you want to run your network on real hardware. First is the connectivity issue, and second is minimizing the number of two-qubit operations.", "_____no_output_____" ], [ "Consider the Foxtail device:", "_____no_output_____" ] ], [ [ "print(cirq.google.Foxtail)", "(0, 0)───(0, 1)───(0, 2)───(0, 3)───(0, 4)───(0, 5)───(0, 6)───(0, 7)───(0, 8)───(0, 9)───(0, 10)\n│ │ │ │ │ │ │ │ │ │ │\n│ │ │ │ │ │ │ │ │ │ │\n(1, 0)───(1, 1)───(1, 2)───(1, 3)───(1, 4)───(1, 5)───(1, 6)───(1, 7)───(1, 8)───(1, 9)───(1, 10)\n" ] ], [ [ "The qubits are arranged in two rows of eleven qubits each, and qubits can only communicate to their nearest neighbors along the horizontal and vertial connections. That does not mesh well with the QNN we designed, where all of the data qubits need to interact with the readout qubit.", "_____no_output_____" ], [ "There is no *in-principle* restriction on the kinds of algorithms you are allowed to run. The solution to the connectivity problem is to make use of SWAP gates, which have the effect of exchanging the states of two (neighboring) qubits. It's equivalent to what you would get if you physically exchanged the positions of two of the qubits in the grid. The problem is that each SWAP operation is costly, so you want to avoid SWAPing as much as possible. We need to think carefully about our algorithm design to minimize the number of SWAPs performed as the circuit is executed.\n\n__Question__: How should we modify our QNN circuit so that it can runs efficiently on the Foxtail device?", "_____no_output_____" ], [ "#### Solution", "_____no_output_____" ], [ "One strategy is to move the readout qubit around as it talks to the other qubits. Suppose the readout qubit starts in the $(0,0)$ position. First it can interact with the qubits in the $(1,0)$ and $(0,1)$ positons like normal, then SWAP with the $(0,1)$ qubit. Now the readout qubit is in the $(0,1)$ position and can interact with the $(1,1)$ and $(0,2)$ qubits before SWAPing with the $(0,2)$ qubit. It continues down the line in this fashion.\n\nLet's code up this circuit:", "_____no_output_____" ] ], [ [ "qnn_fox = cirq.Circuit()\n\nw = 0.2 # Want an explicit numerical weight for later \nfor i in range(10):\n qnn_fox.append([ZXGate(w).on(cirq.GridQubit(1,i), cirq.GridQubit(0,i)),\n ZXGate(w).on(cirq.GridQubit(0,i+1), cirq.GridQubit(0,i)),\n cirq.SWAP(cirq.GridQubit(0,i), cirq.GridQubit(0,i+1))])\n \nqnn_fox.append(ZXGate(w).on(cirq.GridQubit(1,10), cirq.GridQubit(0,10)))\n\nqnn_fox.append([(cirq.S**-1)(cirq.GridQubit(0,10)),cirq.H(cirq.GridQubit(0,10)),\n cirq.measure(cirq.GridQubit(0,10))])\n\nprint(qnn_fox)", "(0, 0): ────X───────X───────×──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────\n │ │ │\n(0, 1): ────┼───────Z^0.2───×───X───────X───────×──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────\n │ │ │ │\n(0, 2): ────┼───────────────────┼───────Z^0.2───×───X───────X───────×──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────\n │ │ │ │ │\n(0, 3): ────┼───────────────────┼───────────────────┼───────Z^0.2───×───X───────X───────×──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────\n │ │ │ │ │ │\n(0, 4): ────┼───────────────────┼───────────────────┼───────────────────┼───────Z^0.2───×───X───────X───────×──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────\n │ │ │ │ │ │ │\n(0, 5): ────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────Z^0.2───×───X───────X───────×──────────────────────────────────────────────────────────────────────────────────────────────────────────\n │ │ │ │ │ │ │ │\n(0, 6): ────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────Z^0.2───×───X───────X───────×──────────────────────────────────────────────────────────────────────────────────────\n │ │ │ │ │ │ │ │ │\n(0, 7): ────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────Z^0.2───×───X───────X───────×──────────────────────────────────────────────────────────────────\n │ │ │ │ │ │ │ │ │ │\n(0, 8): ────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────Z^0.2───×───X───────X───────×──────────────────────────────────────────────\n │ │ │ │ │ │ │ │ │ │ │\n(0, 9): ────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────Z^0.2───×───X───────X───────×──────────────────────────\n │ │ │ │ │ │ │ │ │ │ │ │\n(0, 10): ───┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────Z^0.2───×───X───────S^-1───H───M───\n │ │ │ │ │ │ │ │ │ │ │\n(1, 0): ────Z^0.2───────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼──────────────────────\n │ │ │ │ │ │ │ │ │ │\n(1, 1): ────────────────────────Z^0.2───────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼──────────────────────\n │ │ │ │ │ │ │ │ │\n(1, 2): ────────────────────────────────────────────Z^0.2───────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼──────────────────────\n │ │ │ │ │ │ │ │\n(1, 3): ────────────────────────────────────────────────────────────────Z^0.2───────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼──────────────────────\n │ │ │ │ │ │ │\n(1, 4): ────────────────────────────────────────────────────────────────────────────────────Z^0.2───────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼──────────────────────\n │ │ │ │ │ │\n(1, 5): ────────────────────────────────────────────────────────────────────────────────────────────────────────Z^0.2───────────────┼───────────────────┼───────────────────┼───────────────────┼───────────────────┼──────────────────────\n │ │ │ │ │\n(1, 6): ────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────Z^0.2───────────────┼───────────────────┼───────────────────┼───────────────────┼──────────────────────\n │ │ │ │\n(1, 7): ────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────Z^0.2───────────────┼───────────────────┼───────────────────┼──────────────────────\n │ │ │\n(1, 8): ────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────Z^0.2───────────────┼───────────────────┼──────────────────────\n │ │\n(1, 9): ────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────Z^0.2───────────────┼──────────────────────\n │\n(1, 10): ───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────Z^0.2──────────────────\n" ] ], [ [ "As coded, this circuit still won't run on the Foxtail device. That's because the gates we've defined are not native gates. Cirq has a built-in method that will convert our gates to Xmon gates (which are native for the Foxtail device) and attempt to optimze the circuit by reducing the total number of gates:", "_____no_output_____" ] ], [ [ "cirq.google.optimized_for_xmon(qnn_fox, new_device=cirq.google.Foxtail, allow_partial_czs=True)", "_____no_output_____" ] ], [ [ "Notice how we were able to pass in the `new_device` argument without getting an error messgae. That means the circuit will run properly on the Foxtail.", "_____no_output_____" ], [ "\n__Question__: We were smart to place the SWAP gates and $ZX$ gates next to each other where possible. Why?\n\n__Question__: Can you see any ways to further optimize this circuit by hand? Hint: not all of the qubits are being measured.", "_____no_output_____" ], [ "#### Solutions", "_____no_output_____" ], [ "* Placing the SWAP and $ZX$ gates next to each other lets the optimizer treat the ocmbination of them as a single gate, which leads to fewer total two-qubit gates. \n\n* The state of any qubit which is not being measured does not matter. In particular, any single-qubit gate acting on a non-measured qubit after the last two-qubit gate acting on that qubit will not affect the state of the measured qubit and so can be dropped.", "_____no_output_____" ], [ "### Exercise: Multiple Weights\n\nInstead of just a single weight, create create a neuron with multiple weights. How will you optimize those weights?", "_____no_output_____" ], [ "### Exercise: Analytic Calculation\n\nBecause we stuck to such a simple example, essentially everything in this notebook can be calculated analytically. Do those calculations.", "_____no_output_____" ], [ "### Exercise: Add More \"Quantum\" Operations\n\nThe neuron we constructed essentially does a classial calculation. You can add more ingredients that make the data processing more \"quantum.\" For example, you can add layers of Hadamard gates in between additional layers of $ZX$ gates. This sort of thing was explored in [Farhi and Neven](https://arxiv.org/abs/1802.06002). Try playing around with it.", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "----\n<img src=\"../../../files/refinitiv.png\" width=\"20%\" style=\"vertical-align: top;\">\n\n# Data Library for Python\n\n----", "_____no_output_____" ], [ "## Content layer - News\nThis notebook demonstrates how to retrieve News.", "_____no_output_____" ], [ "#### Learn more\n\nTo learn more about the Refinitiv Data Library for Python please join the Refinitiv Developer Community. By [registering](https://developers.refinitiv.com/iam/register) and [logging](https://developers.refinitiv.com/content/devportal/en_us/initCookie.html) into the Refinitiv Developer Community portal you will have free access to a number of learning materials like \n [Quick Start guides](https://developers.refinitiv.com/en/api-catalog/refinitiv-data-platform/refinitiv-data-library-for-python/quick-start), \n [Tutorials](https://developers.refinitiv.com/en/api-catalog/refinitiv-data-platform/refinitiv-data-library-for-python/learning), \n [Documentation](https://developers.refinitiv.com/en/api-catalog/refinitiv-data-platform/refinitiv-data-library-for-python/docs)\n and much more.\n\n#### Getting Help and Support\n\nIf you have any questions regarding using the API, please post them on \nthe [Refinitiv Data Q&A Forum](https://community.developers.refinitiv.com/spaces/321/index.html). \nThe Refinitiv Developer Community will be happy to help. ", "_____no_output_____" ], [ "## Set the configuration file location\nFor a better ease of use, you have the option to set initialization parameters of the Refinitiv Data Library in the _refinitiv-data.config.json_ configuration file. This file must be located beside your notebook, in your user folder or in a folder defined by the _RD_LIB_CONFIG_PATH_ environment variable. The _RD_LIB_CONFIG_PATH_ environment variable is the option used by this series of examples. The following code sets this environment variable. ", "_____no_output_____" ] ], [ [ "import os\nos.environ[\"RD_LIB_CONFIG_PATH\"] = \"../../../Configuration\"", "_____no_output_____" ] ], [ [ "## Some Imports to start with", "_____no_output_____" ] ], [ [ "import refinitiv.data as rd\nfrom refinitiv.data.content import news\nfrom datetime import timedelta", "_____no_output_____" ] ], [ [ "## Open the data session\n\nThe open_session() function creates and open sessions based on the information contained in the refinitiv-data.config.json configuration file. Please edit this file to set the session type and other parameters required for the session you want to open.", "_____no_output_____" ] ], [ [ "rd.open_session('platform.rdp')", "_____no_output_____" ] ], [ [ "## Retrieve data", "_____no_output_____" ], [ "### Headlines", "_____no_output_____" ], [ "#### Get headlines", "_____no_output_____" ] ], [ [ "response = news.headlines.Definition(\"Apple\").get_data()\nresponse.data.df", "_____no_output_____" ] ], [ [ "#### Get headlines within a range of dates", "_____no_output_____" ] ], [ [ "response = news.headlines.Definition(\n query=\"Refinitiv\",\n date_from=\"20.03.2021\", \n date_to=timedelta(days=-4), \n count=3\n).get_data()\nresponse.data.df", "_____no_output_____" ] ], [ [ "#### Get a limited number of headlines", "_____no_output_____" ] ], [ [ "response = news.headlines.Definition(query = \"Google\", count = 350).get_data()\nresponse.data.df", "_____no_output_____" ] ], [ [ "### Story", "_____no_output_____" ] ], [ [ "response = news.story.Definition(\"urn:newsml:reuters.com:20211003:nNRAgvhyiu:1\").get_data()\nprint(response.data.story.title, '\\n')\nprint(response.data.story.content)", "Google Doodle marks birthday of Spanish ocean scientist María de los Ángeles Alvariño González \n\nFor best results when printing this announcement, please click on link below:\nhttp://newsfile.refinitiv.com/getnewsfile/v1/story?guid=urn:newsml:reuters.com:20211003:nNRAgvhyiu&default-theme=true\n\n\nThe Google Doodle today (3 OCtober) celebrates the 105th birthday of\nSpanish-American professor and marine research biologist María de los\nÁngeles Alvariño González, who is regarded as one of the most important\nSpanish scientists of all time.\n\nBorn in 1916, her love of natural history began with her father's library and\ndeepened as she pursued coastline oceanography research.\n\nAlthough the Spanish Institute of Oceanography (IEO) only accepted men at the\ntime, her university work impressed the organization so much that they\nappointed her as a marine biologist in 1952.\n\nBased in Vigo, she began her pioneering research on zooplankton, tiny\norganisms that serve as the foundation of the oceanic food chain and\nidentified some species to be the best indicators of ocean health.\n\nIn 1953, the British Council awarded Ángeles Alvariño a fellowship that\nresulted in her becoming the first woman to work as a scientist aboard a\nBritish research vessel.\n\nFollowing several expeditions, she furthered her studies in the United States\nwhere she retired as one of the world's most prestigious marine biologists in\n1987.\n\nDuring her career she discovered 22 new species of zooplankton and published\nover 100 scientific papers. Even today she is the only Spanish scientist of\n1,000 in the \"Encyclopedia of World Scientists,\" and a modern research vessel\nin IEO's fleet bears her name.\n\nRead More\n\nAP News Digest 6:30 a.m.\n(https://www.independent.co.uk/news/world/europe/joe-biden-communist-party-east-african-covid-child-b1931394.html)\n\nLa Palma volcano turns 'much more aggressive' and blows open new fissures\n(https://www.independent.co.uk/news/world/europe/la-palma-volcano-eruption-fissures-b1931367.html)\n\nHow security fears for Rutte revealed power of Dutch drug gangs\n(https://www.independent.co.uk/independentpremium/rutte-netherlands-drug-gangs-gangsters-b1931165.html)\n\n\n\nCopyright © 2021 Independent.co.ukk. All rights reserved.\n" ] ], [ [ "## Close the session", "_____no_output_____" ] ], [ [ "rd.close_session()", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Statistic Inference\n- This is a Python base notebook\n- Using `rpy2` for R functions\n\nWe saw some pattern in EDA, naturally, we would like to see if the different between feature are significantly related to the target.", "_____no_output_____" ], [ "## Import libaries", "_____no_output_____" ] ], [ [ "import rpy2\nimport rpy2.robjects as robjects\nfrom rpy2.robjects.packages import importr", "_____no_output_____" ], [ "%load_ext rpy2.ipython", "_____no_output_____" ], [ "%%R\nlibrary(tidyverse)\nlibrary(broom)\nlibrary(GGally)", "R[write to console]: ── Attaching packages ─────────────────────────────────────── tidyverse 1.3.1 ──\n\nR[write to console]: ✔ ggplot2 3.3.5 ✔ purrr 0.3.4\n✔ tibble 3.1.6 ✔ dplyr 1.0.7\n✔ tidyr 1.1.4 ✔ stringr 1.4.0\n✔ readr 2.1.1 ✔ forcats 0.5.1\n\nR[write to console]: ── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──\n✖ dplyr::filter() masks stats::filter()\n✖ dplyr::lag() masks stats::lag()\n\nR[write to console]: Registered S3 method overwritten by 'GGally':\n method from \n +.gg ggplot2\n\n" ] ], [ [ "## Reading the data CSV\nRead in the data CSV and store it as a pandas dataframe named `spotify_df`. ", "_____no_output_____" ] ], [ [ "%%R\nspotify_df <- read_csv(\"data/spotify_data.csv\")\nhead(spotify_df)", "R[write to console]: New names:\n* `` -> ...1\n\n" ] ], [ [ "## Regression", "_____no_output_____" ], [ "### Data Wrangle\n- Remove `song_title` and `artist` for relationship study by regression. As both of them are neither numerical nor categorical features.", "_____no_output_____" ] ], [ [ "%%R\nspotify_df_num <- spotify_df[2:15]\nhead(spotify_df_num)", "# A tibble: 6 × 14\n acousticness danceability duration_ms energy instrumentalness key liveness\n <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>\n1 0.0102 0.833 204600 0.434 0.0219 2 0.165 \n2 0.199 0.743 326933 0.359 0.00611 1 0.137 \n3 0.0344 0.838 185707 0.412 0.000234 2 0.159 \n4 0.604 0.494 199413 0.338 0.51 5 0.0922\n5 0.18 0.678 392893 0.561 0.512 5 0.439 \n6 0.00479 0.804 251333 0.56 0 8 0.164 \n# … with 7 more variables: loudness <dbl>, mode <dbl>, speechiness <dbl>,\n# tempo <dbl>, time_signature <dbl>, valence <dbl>, target <dbl>\n" ] ], [ [ "## Set up regression model", "_____no_output_____" ], [ "Here, I am interested in determining factors associated with `target`. In particular, I will use a Multiple Linear Regression (MLR) Model to study the relation between `target` and all other features.", "_____no_output_____" ] ], [ [ "%%R\nML_reg <- lm( target ~ ., data = spotify_df_num) |> tidy(conf.int = TRUE)\n\nML_reg<- ML_reg |>\n mutate(Significant = p.value < 0.05) |>\n mutate_if(is.numeric, round, 3)\n\nML_reg", "# A tibble: 14 × 8\n term estimate std.error statistic p.value conf.low conf.high Significant\n <chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <lgl> \n 1 (Interce… -0.313 0.206 -1.52 0.128 -0.717 0.09 FALSE \n 2 acoustic… -0.325 0.055 -5.92 0 -0.433 -0.217 TRUE \n 3 danceabi… 0.415 0.078 5.33 0 0.262 0.568 TRUE \n 4 duration… 0 0 4.08 0 0 0 TRUE \n 5 energy 0.09 0.093 0.974 0.33 -0.092 0.272 FALSE \n 6 instrume… 0.268 0.044 6.05 0 0.181 0.354 TRUE \n 7 key 0.001 0.003 0.334 0.739 -0.005 0.007 FALSE \n 8 liveness 0.098 0.07 1.4 0.162 -0.039 0.236 FALSE \n 9 loudness -0.023 0.005 -4.81 0 -0.033 -0.014 TRUE \n10 mode -0.035 0.022 -1.58 0.113 -0.078 0.008 FALSE \n11 speechin… 0.816 0.121 6.74 0 0.579 1.05 TRUE \n12 tempo 0.001 0 1.95 0.052 0 0.002 FALSE \n13 time_sig… -0.009 0.042 -0.205 0.838 -0.091 0.074 FALSE \n14 valence 0.165 0.051 3.24 0.001 0.065 0.265 TRUE \n" ] ], [ [ "- We can see that a lot of features are statiscally correlated with target. They are listed in the table below.", "_____no_output_____" ] ], [ [ "%%R\nML_reg |>\n filter(Significant == TRUE) |>\n select(term) ", "# A tibble: 7 × 1\n term \n <chr> \n1 acousticness \n2 danceability \n3 duration_ms \n4 instrumentalness\n5 loudness \n6 speechiness \n7 valence \n" ] ], [ [ "### GGpairs\nBelow is the ggpair plots to visual the correlation between different features.", "_____no_output_____" ] ], [ [ "%%R\nggpairs(data = spotify_df_num)", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "# Migrating scripts from Framework Mode to Script Mode\n\nThis notebook focus on how to migrate scripts using Framework Mode to Script Mode. The original notebook using Framework Mode can be find here https://github.com/awslabs/amazon-sagemaker-examples/blob/4c2a93114104e0b9555d7c10aaab018cac3d7c04/sagemaker-python-sdk/tensorflow_distributed_mnist/tensorflow_local_mode_mnist.ipynb", "_____no_output_____" ], [ "### Set up the environment", "_____no_output_____" ] ], [ [ "import os\nimport subprocess\nimport sagemaker\nfrom sagemaker import get_execution_role\n\nsagemaker_session = sagemaker.Session()\n\nrole = get_execution_role()", "_____no_output_____" ] ], [ [ "### Download the MNIST dataset", "_____no_output_____" ] ], [ [ "import utils\nfrom tensorflow.examples.tutorials.mnist import input_data\nimport tensorflow as tf\n\ndata_sets = input_data.read_data_sets('data', dtype=tf.uint8, reshape=False, validation_size=5000)\n\nutils.convert_to(data_sets.train, 'train', 'data')\nutils.convert_to(data_sets.validation, 'validation', 'data')\nutils.convert_to(data_sets.test, 'test', 'data')", "_____no_output_____" ] ], [ [ "### Upload the data\nWe use the ```sagemaker.Session.upload_data``` function to upload our datasets to an S3 location. The return value inputs identifies the location -- we will use this later when we start the training job.", "_____no_output_____" ] ], [ [ "inputs = sagemaker_session.upload_data(path='data', key_prefix='data/mnist')", "_____no_output_____" ] ], [ [ "# Construct an entry point script for training \nOn this example, we assume that you aready have a Framework Mode training script named `mnist.py`:", "_____no_output_____" ] ], [ [ "!pygmentize 'mnist.py'", "_____no_output_____" ] ], [ [ "The training script `mnist.py` include the Framework Mode functions ```model_fn```, ```train_input_fn```, ```eval_input_fn```, and ```serving_input_fn```. We need to create a entrypoint script that uses the functions above to create a ```tf.estimator```:", "_____no_output_____" ] ], [ [ "%%writefile train.py\n\nimport argparse\n# import original framework mode script\nimport mnist\n\nimport tensorflow as tf\n\nif __name__ == '__main__':\n parser = argparse.ArgumentParser()\n\n # read hyperparameters as script arguments\n parser.add_argument('--training_steps', type=int)\n parser.add_argument('--evaluation_steps', type=int)\n\n args, _ = parser.parse_known_args()\n\n # creates a tf.Estimator using `model_fn` that saves models to /opt/ml/model\n estimator = tf.estimator.Estimator(model_fn=mnist.model_fn, model_dir='/opt/ml/model')\n\n\n # creates parameterless input_fn function required by the estimator\n def input_fn():\n return mnist.train_input_fn(training_dir='/opt/ml/input/data/training', params=None)\n\n\n train_spec = tf.estimator.TrainSpec(input_fn, max_steps=args.training_steps)\n\n\n # creates parameterless serving_input_receiver_fn function required by the exporter\n def serving_input_receiver_fn():\n return mnist.serving_input_fn(params=None)\n\n\n exporter = tf.estimator.LatestExporter('Servo',\n serving_input_receiver_fn=serving_input_receiver_fn)\n\n\n # creates parameterless input_fn function required by the evaluation\n def input_fn():\n return mnist.eval_input_fn(training_dir='/opt/ml/input/data/training', params=None)\n\n\n eval_spec = tf.estimator.EvalSpec(input_fn, steps=args.evaluation_steps, exporters=exporter)\n \n # start training and evaluation\n tf.estimator.train_and_evaluate(estimator=estimator, train_spec=train_spec, eval_spec=eval_spec)", "_____no_output_____" ] ], [ [ "## Changes in the SageMaker TensorFlow estimator\n\nWe need to create a TensorFlow estimator pointing to ```train.py``` as the entrypoint:", "_____no_output_____" ] ], [ [ "from sagemaker.tensorflow import TensorFlow\n\nmnist_estimator = TensorFlow(entry_point='train.py',\n dependencies=['mnist.py'],\n role='SageMakerRole',\n framework_version='1.13',\n hyperparameters={'training_steps':10, 'evaluation_steps':10},\n py_version='py3',\n train_instance_count=1,\n train_instance_type='local')\n\nmnist_estimator.fit(inputs)", "_____no_output_____" ] ], [ [ "# Deploy the trained model to prepare for predictions\n\nThe deploy() method creates an endpoint (in this case locally) which serves prediction requests in real-time.", "_____no_output_____" ] ], [ [ "mnist_predictor = mnist_estimator.deploy(initial_instance_count=1, instance_type='local')", "_____no_output_____" ] ], [ [ "# Invoking the endpoint", "_____no_output_____" ] ], [ [ "import numpy as np\nfrom tensorflow.examples.tutorials.mnist import input_data\n\nmnist = input_data.read_data_sets(\"/tmp/data/\", one_hot=True)\n\nfor i in range(10):\n data = mnist.test.images[i].tolist()\n\n predict_response = mnist_predictor.predict(data)\n \n print(\"========================================\")\n label = np.argmax(mnist.test.labels[i])\n print(\"label is {}\".format(label))\n print(\"prediction is {}\".format(predict_response))", "_____no_output_____" ] ], [ [ "# Clean-up\n\nDeleting the local endpoint when you're finished is important since you can only run one local endpoint at a time.", "_____no_output_____" ] ], [ [ "mnist_estimator.delete_endpoint()", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "CC-BY-4.0" ]

[ "CC-BY-4.0" ]

[ "CC-BY-4.0" ]

[ [ [ "# -*- coding: utf-8 -*-\nimport sys\nimport os\nimport numpy as np\nimport pandas as pd\nimport cobra\n\nprint('Python version:', sys.version)\nprint('numpy version:', np.__version__)\nprint('pandas version:', pd.__version__)\nprint('cobrapy version:', cobra.__version__)", "Python version: 3.7.3 (default, Mar 27 2019, 17:13:21) [MSC v.1915 64 bit (AMD64)]\nnumpy version: 1.16.4\npandas version: 1.0.3\ncobrapy version: 0.15.4\n" ], [ "def AddRxn(model, newRxnFile):\n \"\"\"Function of adding new reactions to the model.\"\"\"\n n1 = len(model.reactions)\n AllAddRxn = pd.read_csv(newRxnFile, sep=',', index_col='RxnID', skipinitialspace=True)\n n2 = len(AllAddRxn)\n for i in range(n2):\n ID = AllAddRxn.index.values[i]\n addRxn = cobra.Reaction(ID)\n model.add_reactions([addRxn])\n addRxnInf = model.reactions[n1 + i]\n addRxnInf.name = AllAddRxn.loc[ID, 'RxnName']\n addRxnInf.reaction = AllAddRxn.loc[ID, 'RxnFormula']\n addRxnInf.subsystem = AllAddRxn.loc[ID, 'Subsystem']\n addRxnInf.lower_bound = AllAddRxn.loc[ID, 'LowerBound']\n addRxnInf.upper_bound = AllAddRxn.loc[ID, 'UpperBound']\n return model\n", "_____no_output_____" ], [ "def flux2file(model, product, psw, output_dir='tmp'):\n \"\"\"Function of exporting flux data.\"\"\"\n n = len(model.reactions)\n modelMatrix = np.empty([n, 9], dtype = object)\n for i in range(len(model.reactions)):\n x = model.reactions[i]\n modelMatrix[i, 0] = i + 1\n modelMatrix[i, 1] = x.id\n modelMatrix[i, 2] = x.name\n modelMatrix[i, 3] = x.reaction\n modelMatrix[i, 4] = x.subsystem\n modelMatrix[i, 5] = x.lower_bound\n modelMatrix[i, 6] = x.upper_bound\n modelMatrix[i, 7] = x.flux\n modelMatrix[i, 8] = abs(x.flux)\n \n df = pd.DataFrame(data = modelMatrix, \n columns = ['N', 'RxnID', 'RxnName', 'Reaction', 'SubSystem', \n 'LowerBound', 'UpperBound', 'Flux-core', 'abs(Flux)'])\n if not os.path.exists(output_dir):\n os.mkdir(output_dir)\n filepath = os.path.join(output_dir, '{}_{}.xlsx'.format(product, psw))\n df.to_excel(filepath, index=False)\n ", "_____no_output_____" ], [ "model = cobra.Model()\n\nAddRxn(model,'CBBrxns.csv')\n\n# the model has a constraint of producing 1 of 3pg (sink_3pg)\n# set the objective to minimize ATP cost\nmodel.objective = {model.reactions.DM_atp: 1}\nmodel.objective_direction = 'min'\n\n# set the carboxylation and oxygenation reaction ratio \n# of RuBisCO to be 3:1\nrubisco_flux = model.problem.Constraint(\n model.reactions.RBPC.flux_expression - 3 * model.reactions.RBPO.flux_expression,\n lb = 0, \n ub = 0\n)\nmodel.add_cons_vars(rubisco_flux)", "unknown metabolite 'co2' created\nunknown metabolite 'h2o' created\nunknown metabolite 'rb15bp' created\nunknown metabolite '3pg' created\nunknown metabolite 'h' created\nunknown metabolite 'o2' created\nunknown metabolite '2pglyc' created\nunknown metabolite 'atp' created\nunknown metabolite '13dpg' created\nunknown metabolite 'adp' created\nunknown metabolite 'nadph' created\nunknown metabolite 'g3p' created\nunknown metabolite 'nadp' created\nunknown metabolite 'pi' created\nunknown metabolite 'dhap' created\nunknown metabolite 'fdp' created\nunknown metabolite 'f6p' created\nunknown metabolite 'e4p' created\nunknown metabolite 'xu5p' created\nunknown metabolite 's17bp' created\nunknown metabolite 's7p' created\nunknown metabolite 'r5p' created\nunknown metabolite 'ru5p' created\nunknown metabolite 'amp' created\nunknown metabolite 'nad' created\nunknown metabolite 'nadh' created\nunknown metabolite 'fdxo' created\nunknown metabolite 'fdxrd' created\nunknown metabolite 'ppi' created\nunknown metabolite 'hco3' created\nunknown metabolite 'nh3' created\n" ], [ "photores = {\n 'NPR': 'a_NPRrxns.csv', # natural photorespiration\n 'GLC': 'b_GLCrxns.csv', # glycerate bypass\n 'OX': 'c_OXrxns.csv', # glycolate oxidation pathway\n 'A5P': 'd_A5Prxns.csv', # arabinose-5-phosphate shunt\n '3OHP': 'e_3OHPrxns.csv', # 3-hydroxypropionate bypass\n 'TACO': 'f_TACOrxns.csv', # tartronyl-CoA pathway\n}\n\ncost_df = pd.DataFrame()", "_____no_output_____" ], [ "for psw, rxns in photores.items():\n with model as m:\n AddRxn(m, rxns)\n m.optimize()\n flux2file(m,'3pg',psw,'output')\n for cost in ['DM_atp', 'DM_e', 'Fdr', 'EX_co2', 'RBPC', 'RBPO']:\n cost_df.loc[cost, psw] = abs(m.reactions.get_by_id(cost).flux)\n ", "unknown metabolite 'glyclt' created\nunknown metabolite 'glx' created\nunknown metabolite 'h2o2' created\nunknown metabolite 'ser' created\nunknown metabolite 'gly' created\nunknown metabolite 'hpyr' created\nunknown metabolite 'glu' created\nunknown metabolite 'akg' created\nunknown metabolite 'gln' created\nunknown metabolite 'mlthf' created\nunknown metabolite 'thf' created\nunknown metabolite 'glyc' created\nunknown metabolite 'glyclt' created\nunknown metabolite 'glx' created\nunknown metabolite '2h3oppan' created\nunknown metabolite 'glyc' created\nunknown metabolite 'glyclt' created\nunknown metabolite 'glx' created\nunknown metabolite 'h2o2' created\nunknown metabolite 'accoa' created\nunknown metabolite 'mal' created\nunknown metabolite 'coa' created\nunknown metabolite 'pyr' created\nunknown metabolite 'glyclt' created\nunknown metabolite 'coa' created\nunknown metabolite 'glyccoa' created\nunknown metabolite 'gcald' created\nunknown metabolite 'ara5p' created\nunknown metabolite 'glyclt' created\nunknown metabolite 'glx' created\nunknown metabolite 'ppcoa' created\nunknown metabolite 'mmcoa' created\nunknown metabolite 'citmcoa' created\nunknown metabolite 'pyr' created\nunknown metabolite 'accoa' created\nunknown metabolite 'malcoa' created\nunknown metabolite '3hpp' created\nunknown metabolite 'coa' created\nunknown metabolite 'pep' created\nunknown metabolite '2pg' created\nunknown metabolite 'glyclt' created\nunknown metabolite 'coa' created\nunknown metabolite 'glyccoa' created\nunknown metabolite 'tarcoa' created\nunknown metabolite '2h3oppan' created\nunknown metabolite 'glyc' created\n" ], [ "# Assuming we cannot improve the GCC hydrolysis\n# The GCC M5 hydrolyse 3.9 ATP per carboxylation (Fig. 2c)\nwith model as m:\n AddRxn(m, 'f_TACOrxns.csv')\n m.reactions.GCC.add_metabolites({'atp': -2.9, 'adp': 2.9, 'pi': 2.9})\n m.optimize()\n flux2file(m,'3pg', 'TACO_2','output')\n for cost in ['DM_atp', 'DM_e', 'Fdr', 'EX_co2', 'RBPC', 'RBPO']:\n cost_df.loc[cost, 'TACO_2'] = abs(m.reactions.get_by_id(cost).flux)", "unknown metabolite 'glyclt' created\nunknown metabolite 'coa' created\nunknown metabolite 'glyccoa' created\nunknown metabolite 'tarcoa' created\nunknown metabolite '2h3oppan' created\nunknown metabolite 'glyc' created\n" ], [ "cost_df.to_excel('TaCo costs comparison.xlsx')\ncost_df", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "The Linnerud dataset is a multi-output regression dataset. It consists of three excercise (data) and three physiological (target) variables collected from twenty middle-aged men in a fitness club:\n\nphysiological - CSV containing 20 observations on 3 physiological variables:\nWeight, Waist and Pulse.\n\nexercise - CSV containing 20 observations on 3 exercise variables:\nChins, Situps and Jumps.", "_____no_output_____" ], [ "To create a Regression model that would plot the relationship between the waistline and how many situps are accomplished we need to take both of the variables into separate 1 dimensional vectors, divide them into training and testing sets, train a linear regression model using train set and check its output on the test set", "_____no_output_____" ] ], [ [ "import matplotlib.pyplot as plt\nimport numpy as np\nfrom sklearn import datasets, linear_model, model_selection", "_____no_output_____" ], [ "X, y = datasets.load_linnerud(return_X_y=True)\nprint(X.shape)\nprint(X[0])", "(20, 3)\n[ 5. 162. 60.]\n" ] ], [ [ "Taking Waist and Situps variables:", "_____no_output_____" ] ], [ [ "X = X[:, np.newaxis, 1]\ny = y[:, np.newaxis, 1]", "_____no_output_____" ], [ "print(X.shape)\nprint(y.shape)", "(20, 1)\n(20, 1)\n" ], [ "X_train, X_test, y_train, y_test = model_selection.train_test_split(X, y, test_size=0.33)", "_____no_output_____" ], [ "model = linear_model.LinearRegression()\nmodel.fit(X_train, y_train)", "_____no_output_____" ], [ "y_pred = model.predict(X_test)", "_____no_output_____" ], [ "plt.scatter(X_test, y_test, color='black')\nplt.plot(X_test, y_pred, color='blue', linewidth=3)\nplt.show()", "_____no_output_____" ] ], [ [ "We can see that there is an inverse correlation between number of situps someone does and the waistline", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code" ], [ "markdown" ] ]

[ "Unlicense" ]

[ "Unlicense" ]

[ "Unlicense" ]

[ [ [ "import numpy as np\nimport torch\nimport os\nimport pickle\nimport matplotlib.pyplot as plt\n# %matplotlib inline\n# plt.rcParams['figure.figsize'] = (20, 20)\n# plt.rcParams['image.interpolation'] = 'bilinear'\n\nimport sys\nsys.path.append('../train/')\n\nfrom torch.autograd import Variable\nfrom torch.utils.data import DataLoader, Dataset\nimport torchvision.datasets as datasets\nimport torchvision\nimport torchvision.transforms as T\nimport torch.nn.functional as F\nimport torch.nn as nn\n\nimport collections\nimport numbers\nimport random\nimport math\nfrom PIL import Image, ImageOps, ImageEnhance\nimport time\nfrom torch.utils.data import Dataset\n\nfrom networks.DilatedSegUNet_new_version import DilatedSegUNet_new_version\nimport tool\nfrom tqdm import tqdm\n\nflip_index = ['16', '15', '14', '13', '12', '11', '10']", "_____no_output_____" ], [ "NUM_CHANNELS = 3\nNUM_CLASSES = 2 \nBATCH_SIZE = 2\nW, H = 1918, 1280\nSTRIDE = 512\nIMAGE_SIZE = 1024\ntest_mask_path = '../../data/test_masks/DilatedSegUNet_new_version/'\nweight_path = '../_weights/DilatedSegUNet_new_version-fold0-end.pth'", "_____no_output_____" ], [ "def load_model(filename, model):\n checkpoint = torch.load(filename)\n model.load_state_dict(checkpoint['model_state'])", "_____no_output_____" ], [ "model = DilatedSegUNet_new_version()\nmodel = model.cuda()\nmodel.eval()\nload_model(weight_path, model)", "_____no_output_____" ], [ "test_path = '../../data/images/test/'\n\nif not os.path.exists(test_mask_path):\n os.makedirs(test_mask_path)", "_____no_output_____" ], [ "test_names = os.listdir(test_path)\ntest_names = sorted(test_names)", "_____no_output_____" ], [ "coords_full = np.zeros((BATCH_SIZE, 2, H, W), dtype='float32')\nxx,yy = np.meshgrid(np.linspace(-0.5,0.5,W), np.linspace(-0.5,0.5,H))\ncoord = np.concatenate([xx[np.newaxis,...], yy[np.newaxis,...]],0).astype('float32')\nfor i in range(BATCH_SIZE):\n coords_full[i] = coord", "_____no_output_____" ], [ "with torch.no_grad():\n batch_size = BATCH_SIZE\n normalize_mean = [.485, .456, .406]\n normalize_std = [.229, .224, .225]\n\n test_names = sorted(os.listdir(test_path))\n for image_pack in tqdm(range(len(test_names) // batch_size)):\n images = np.zeros((batch_size, 3, H, W), dtype='float32')\n test_masks = np.zeros((batch_size, 2, H, W), dtype='float32')\n ifflip = [False] * batch_size\n image_batch_names = test_names[image_pack * batch_size: image_pack * batch_size + batch_size]\n mask_names = [input_name.split('.')[0] + '.png' for input_name in image_batch_names]\n \n for idx, image_name in enumerate(image_batch_names):\n image = Image.open(os.path.join(test_path, image_name))\n angle = image_name.split('.')[0].split('_')[-1]\n if angle in flip_index:\n ifflip[idx] = True\n image = ImageOps.mirror(image)\n\n image = np.array(image).astype('float') / 255\n image = image.transpose(2, 0, 1)\n\n for i in range(3):\n image[i] = (image[i] - normalize_mean[i]) / normalize_std[i]\n\n images[idx] = image\n \n\n\n for h_idx in range(int(math.ceil((H - STRIDE) / STRIDE))):\n h_start = h_idx * STRIDE\n h_end = h_start + IMAGE_SIZE\n if h_end > H:\n h_end = H\n h_start = h_end - IMAGE_SIZE\n for w_idx in range(int(math.ceil((W - STRIDE) / STRIDE))):\n w_start = w_idx * STRIDE\n w_end = w_start + IMAGE_SIZE\n if w_end > W:\n w_end = W\n w_start = w_end - IMAGE_SIZE\n\n input_batchs = images[:, :, h_start:h_end, w_start:w_end]\n input_tensor = torch.from_numpy(input_batchs).cuda()\n inputs = Variable(input_tensor)\n \n coord_batchs = coords_full[:, :, h_start:h_end, w_start:w_end]\n coord_batchs = coord_batchs[:, :, ::16, ::16]\n coord_tensor = torch.from_numpy(coord_batchs).cuda()\n coords = Variable(coord_tensor)\n \n outputs = model(inputs, coords)\n ouputs = outputs.cpu().data.numpy()\n\n test_masks[:, :, h_start:h_end, w_start:w_end] += ouputs\n \n test_masks = np.argmax(test_masks, axis=1).astype('uint8')\n for idx in range(batch_size):\n output_PIL = Image.fromarray(test_masks[idx].astype('uint8')*255).convert('1')\n if ifflip[idx]:\n output_PIL = ImageOps.mirror(output_PIL)\n mask_name = mask_names[idx]\n output_PIL.save(test_mask_path + mask_name)\n", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "⊕ [sanskrit-coders/indic_transliteration: Python package for indic script transliteration](https://github.com/sanskrit-coders/indic_transliteration)\n\n```\npip install indic_transliteration\npip install git+https://github.com/sanskrit-coders/indic_transliteration/@master \n```", "_____no_output_____" ] ], [ [ "from indic_transliteration import sanscript\nfrom indic_transliteration.sanscript import SchemeMap, SCHEMES, transliterate\n\ndata = 'idam adbhutam'\nprint(transliterate(data, sanscript.HK, sanscript.TELUGU))", "ఇదమ్ అద్భుతమ్\n" ], [ "print(transliterate(data, sanscript.ITRANS, sanscript.DEVANAGARI))", "इदम् अद्भुतम्\n" ], [ "scheme_map = SchemeMap(SCHEMES[sanscript.VELTHUIS], SCHEMES[sanscript.TELUGU])\nprint(transliterate(data, scheme_map=scheme_map))", "ఇదమ్ అద్భుతమ్\n" ], [ "# Dravidian language extension (印度南部达罗毗荼人)\nfrom indic_transliteration import xsanscript\n# from indic_transliteration.xsanscript import SchemeMap, SCHEMES, transliterate\n\ndata = 'असय औषधिः ग्रन्थः। ऎ ऒ यॆक्ककॊ?'\nprint(transliterate(data, xsanscript.DEVANAGARI, xsanscript.KANNADA))", "ಅಸಯ ಔಷಧಿಃ ಗ್ರನ್ಥಃ। ಎ ಒ ಯೆಕ್ಕಕೊ?\n" ] ] ]

[ "markdown", "code" ]

[ [ "markdown" ], [ "code", "code", "code", "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "<a href=\"https://colab.research.google.com/github/yushan111/analytics-zoo/blob/add-autoestimator-quick-start/docs/docs/colab-notebook/orca/quickstart/autoestimator_pytorch_lenet_mnist.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>", "_____no_output_____" ], [ "\n---", "_____no_output_____" ], [ "##### Copyright 2018 Analytics Zoo Authors.", "_____no_output_____" ] ], [ [ "#@title Licensed under the Apache License, Version 2.0 (the \"License\");\n# you may not use this file except in compliance with the License.\n# You may obtain a copy of the License at\n#\n# http://www.apache.org/licenses/LICENSE-2.0\n#\n# Unless required by applicable law or agreed to in writing, software\n# distributed under the License is distributed on an \"AS IS\" BASIS,\n# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n# See the License for the specific language governing permissions and\n# limitations under the License.\n#", "_____no_output_____" ] ], [ [ "## **Environment Preparation**", "_____no_output_____" ], [ "**Install Java 8**\n\nRun the cell on the **Google Colab** to install jdk 1.8.\n\n**Note:** if you run this notebook on your computer, root permission is required when running the cell to install Java 8. (You may ignore this cell if Java 8 has already been set up in your computer).", "_____no_output_____" ] ], [ [ "# Install jdk8\n!apt-get install openjdk-8-jdk-headless -qq > /dev/null\nimport os\n# Set environment variable JAVA_HOME.\nos.environ[\"JAVA_HOME\"] = \"/usr/lib/jvm/java-8-openjdk-amd64\"\n!update-alternatives --set java /usr/lib/jvm/java-8-openjdk-amd64/jre/bin/java\n!java -version", "_____no_output_____" ] ], [ [ "**Install Analytics Zoo**\n\n ", "_____no_output_____" ], [ "[Conda](https://docs.conda.io/projects/conda/en/latest/user-guide/install/) is needed to prepare the Python environment for running this example. \n\n**Note**: The following code cell is specific for setting up conda environment on Colab; for general conda installation, please refer to the [install guide](https://docs.conda.io/projects/conda/en/latest/user-guide/install/) for more details.", "_____no_output_____" ] ], [ [ "import sys\n\n# Get current python version\nversion_info = sys.version_info\npython_version = f\"{version_info.major}.{version_info.minor}.{version_info.micro}\"", "_____no_output_____" ], [ "# Install Miniconda\n!wget https://repo.continuum.io/miniconda/Miniconda3-4.5.4-Linux-x86_64.sh\n!chmod +x Miniconda3-4.5.4-Linux-x86_64.sh\n!./Miniconda3-4.5.4-Linux-x86_64.sh -b -f -p /usr/local\n\n# Update Conda\n!conda install --channel defaults conda python=$python_version --yes\n!conda update --channel defaults --all --yes\n\n# Append to the sys.path\n_ = (sys.path\n .append(f\"/usr/local/lib/python{version_info.major}.{version_info.minor}/site-packages\"))\n\nos.environ['PYTHONHOME']=\"/usr/local\"", "_____no_output_____" ] ], [ [ "You can install the latest pre-release version using `pip install --pre --upgrade analytics-zoo[ray]`.", "_____no_output_____" ] ], [ [ "# Install latest pre-release version of Analytics Zoo \n# Installing Analytics Zoo from pip will automatically install pyspark, bigdl, and their dependencies.\n!pip install --pre --upgrade analytics-zoo[ray]", "_____no_output_____" ], [ "# Install python dependencies\n!pip install torch==1.7.1 torchvision==0.8.2", "_____no_output_____" ] ], [ [ "## **Automated hyper-parameter search for PyTorch using Orca APIs**\n\nIn this guide we will describe how to enable automated hyper-parameter search for PyTorch using Orca `AutoEstimator` in 5 simple steps.", "_____no_output_____" ] ], [ [ "# import necesary libraries and modules\nfrom __future__ import print_function\nimport os\nimport argparse\n\nfrom zoo.orca import init_orca_context, stop_orca_context\nfrom zoo.orca import OrcaContext", "_____no_output_____" ] ], [ [ "### **Step 1: Init Orca Context**", "_____no_output_____" ] ], [ [ "# recommended to set it to True when running Analytics Zoo in Jupyter notebook. \nOrcaContext.log_output = True # (this will display terminal's stdout and stderr in the Jupyter notebook).\n\ncluster_mode = \"local\"\n\nif cluster_mode == \"local\":\n init_orca_context(cores=4, memory=\"2g\", init_ray_on_spark=True) # run in local mode\nelif cluster_mode == \"k8s\":\n init_orca_context(cluster_mode=\"k8s\", num_nodes=2, cores=4, init_ray_on_spark=True) # run on K8s cluster\nelif cluster_mode == \"yarn\":\n init_orca_context(\n cluster_mode=\"yarn-client\", cores=4, num_nodes=2, memory=\"2g\", init_ray_on_spark=True, \n driver_memory=\"10g\", driver_cores=1) # run on Hadoop YARN cluster", "_____no_output_____" ] ], [ [ "This is the only place where you need to specify local or distributed mode. View [Orca Context](https://analytics-zoo.readthedocs.io/en/latest/doc/Orca/Overview/orca-context.html) for more details.\n\n**Note**: You should export HADOOP_CONF_DIR=/path/to/hadoop/conf/dir when you run on Hadoop YARN cluster.", "_____no_output_____" ], [ "### **Step 2: Define the Model**\nYou may define your model, loss and optimizer in the same way as in any standard PyTorch program.", "_____no_output_____" ] ], [ [ "import torch\nimport torch.nn as nn\nimport torch.nn.functional as F\n\nclass LeNet(nn.Module):\n def __init__(self, fc1_hidden_size=500):\n super(LeNet, self).__init__()\n self.conv1 = nn.Conv2d(1, 20, 5, 1)\n self.conv2 = nn.Conv2d(20, 50, 5, 1)\n self.fc1 = nn.Linear(4*4*50, fc1_hidden_size)\n self.fc2 = nn.Linear(fc1_hidden_size, 10)\n\n def forward(self, x):\n x = F.relu(self.conv1(x))\n x = F.max_pool2d(x, 2, 2)\n x = F.relu(self.conv2(x))\n x = F.max_pool2d(x, 2, 2)\n x = x.view(-1, 4*4*50)\n x = F.relu(self.fc1(x))\n x = self.fc2(x)\n return F.log_softmax(x, dim=1)\n\ncriterion = nn.NLLLoss()", "_____no_output_____" ] ], [ [ "After defining your model, you need to define a *Model Creator Function* that returns an instance of your model, and a *Optimizer Creator Function* that returns a PyTorch optimizer. Note that both the *Model Creator Function* and the *Optimizer Creator Function* should take `config` as input and get the hyper-parameter values from `config`.", "_____no_output_____" ] ], [ [ "def model_creator(config):\n model = LeNet(fc1_hidden_size=config[\"fc1_hidden_size\"])\n return model\n\ndef optim_creator(model, config):\n return torch.optim.Adam(model.parameters(), lr=config[\"lr\"])", "_____no_output_____" ] ], [ [ "### **Step 3: Define Dataset**\n\nYou can define the train and validation datasets using *Data Creator Function* that has one parameter of `config` and returns a PyTorch `DataLoader`.", "_____no_output_____" ] ], [ [ "import torch\nfrom torchvision import datasets, transforms\n\ntorch.manual_seed(0)\ndir = './dataset'\ntest_batch_size = 640\n\ndef train_loader_creator(config):\n train_loader = torch.utils.data.DataLoader(\n datasets.MNIST(dir, train=True, download=True,\n transform=transforms.Compose([\n transforms.ToTensor(),\n transforms.Normalize((0.1307,), (0.3081,))\n ])),\n batch_size=config[\"batch_size\"], shuffle=True)\n return train_loader\n\ndef test_loader_creator(config):\n test_loader = torch.utils.data.DataLoader(\n datasets.MNIST(dir, train=False, download=True,\n transform=transforms.Compose([\n transforms.ToTensor(),\n transforms.Normalize((0.1307,), (0.3081,))\n ])),\n batch_size=test_batch_size, shuffle=False)\n return test_loader", "_____no_output_____" ] ], [ [ "### **Step 4: Define search space**\n\nYou should define a dictionary as your hyper-parameter search space. The keys are hyper-parameter names which should be the same with those in your creators, and you can specify how you want to sample each hyper-parameter in the values of the search space. See more about [automl.hp](https://analytics-zoo.readthedocs.io/en/latest/doc/PythonAPI/AutoML/automl.html#orca-automl-hp).\n", "_____no_output_____" ] ], [ [ "from zoo.orca.automl import hp\n\nsearch_space = {\n \"fc1_hidden_size\": hp.choice([500, 600]),\n \"lr\": hp.choice([0.001, 0.003]),\n \"batch_size\": hp.choice([160, 320, 640]),\n}", "_____no_output_____" ] ], [ [ "### **Step 5: Automatically fit and search *with* Orca AutoEstimator**", "_____no_output_____" ], [ "First, create an AutoEstimator. You can refer to [AutoEstimator API doc](https://analytics-zoo.readthedocs.io/en/latest/doc/PythonAPI/AutoML/automl.html#orca-automl-auto-estimator) for more details.", "_____no_output_____" ] ], [ [ "from zoo.orca.automl.auto_estimator import AutoEstimator\n\nauto_est = AutoEstimator.from_torch(model_creator=model_creator,\n optimizer=optim_creator,\n loss=criterion,\n logs_dir=\"/tmp/zoo_automl_logs\",\n resources_per_trial={\"cpu\": 2},\n name=\"lenet_mnist\")", "_____no_output_____" ] ], [ [ "Next, use the auto estimator to fit and search for the best hyper-parameter set.", "_____no_output_____" ] ], [ [ "auto_est.fit(data=train_loader_creator,\n validation_data=test_loader_creator,\n search_space=search_space,\n n_sampling=2,\n epochs=1,\n metric=\"accuracy\")", "_____no_output_____" ] ], [ [ "Finally, you can get the best learned model and the best hyper-parameters.", "_____no_output_____" ] ], [ [ "best_model = auto_est.get_best_model().model # will change later", "_____no_output_____" ], [ "best_config = auto_est.get_best_config()\nprint(best_config)", "_____no_output_____" ] ], [ [ "You can use the best learned model and the best hyper-parameters as you want. Here, we demonstrate how to evaluate on the test dataset.", "_____no_output_____" ] ], [ [ "test_loader = test_loader_creator(best_config)\nbest_model.eval()\ncorrect = 0\nwith torch.no_grad():\n for data, target in test_loader:\n output = best_model(data)\n pred = output.data.max(1, keepdim=True)[1]\n correct += pred.eq(target.data.view_as(pred)).sum().numpy()\naccuracy = 100. * correct / len(test_loader.dataset)\nprint(f\"accuracy is {accuracy}%\")", "_____no_output_____" ] ], [ [ "You can find the accuracy of the best model has reached 98%.", "_____no_output_____" ] ], [ [ "# stop orca context when program finishes\nstop_orca_context()", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "CC-BY-4.0", "BSD-3-Clause" ]

[ "CC-BY-4.0", "BSD-3-Clause" ]

[ "CC-BY-4.0", "BSD-3-Clause" ]

[ [ [ "<a href=\"https://colab.research.google.com/github/NeuromatchAcademy/course-content-dl/blob/main/tutorials/W1D2_LinearDeepLearning/student/W1D2_Tutorial1.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>", "_____no_output_____" ], [ "\n# DL Neuromatch Academy: Week 1, Day 2, Tutorial 1\n# Gradients and AutoGrad\n\n__Content creators:__ Saeed Salehi, Vladimir Haltakov, Andrew Saxe\n\n\n\n__Content reviewers:__ Polina Turishcheva, Atnafu Lambebo, Yu-Fang Yang\n\n__Content editors:__ Anoop Kulkarni\n\n__Production editors:__ Khalid Almubarak, Spiros Chavlis", "_____no_output_____" ], [ "---\n#Tutorial Objectives\n\nDay 2 Tutorial 1 will continue on buiding PyTorch skillset and motivate its core functionality, Autograd. In this notebook, we will cover the key concepts and ideas of:\n\n* Gradient descent\n* PyTorch Autograd\n* PyTorch nn module\n\n\n", "_____no_output_____" ] ], [ [ "#@markdown Tutorial slides\n# you should link the slides for all tutorial videos here (we will store pdfs on osf)\n\nfrom IPython.display import HTML\nHTML('<iframe src=\"https://docs.google.com/presentation/d/1kfWWYhSIkczYfjebhMaqQILTCu7g94Q-o_ZcWb1QAKs/embed?start=false&loop=false&delayms=3000\" frameborder=\"0\" width=\"960\" height=\"569\" allowfullscreen=\"true\" mozallowfullscreen=\"true\" webkitallowfullscreen=\"true\"></iframe>')", "_____no_output_____" ] ], [ [ "---\n# Setup\n", "_____no_output_____" ] ], [ [ "# Imports\n\nimport numpy as np\nimport matplotlib.pyplot as plt\nimport torch\nfrom torch import nn\nimport time\nimport random\n\nfrom tqdm.notebook import tqdm, trange", "_____no_output_____" ], [ "#@title Figure settings\n\nimport ipywidgets as widgets # interactive display\n%config InlineBackend.figure_format = 'retina'\nplt.style.use(\"https://raw.githubusercontent.com/NeuromatchAcademy/course-content/master/nma.mplstyle\")", "_____no_output_____" ], [ "#@title Plotting functions\n\ndef ex3_plot(epochs, losses, values, gradients):\n f, (plot1, plot2, plot3) = plt.subplots(3, 1, sharex=True, figsize=(10, 7))\n plot1.set_title(\"Cross Entropy Loss\")\n plot1.plot(np.linspace(1, epochs, epochs), losses, color='r')\n\n plot2.set_title(\"First Parameter value\")\n plot2.plot(np.linspace(1, epochs, epochs), values, color='c')\n\n plot3.set_title(\"First Parameter gradient\")\n plot3.plot(np.linspace(1, epochs, epochs), gradients, color='m')\n plot3.set_xlabel(\"Epoch\")\n plt.show()", "_____no_output_____" ], [ "#@title Helper functions\n\nseed = 1943 # McCulloch & Pitts (1943)\nrandom.seed(seed)\ntorch.manual_seed(seed)\ntorch.cuda.manual_seed_all(seed)\ntorch.cuda.manual_seed(seed)\nnp.random.seed(seed)\ntorch.backends.cudnn.deterministic = True\ntorch.backends.cudnn.benchmark = False", "_____no_output_____" ] ], [ [ "---\n# Section 1: Gradient Descent Algorithm\n\nSince the goal of most learning algorithms is **minimizing the risk (cost) function**, optimization is the soul of learning! The gradient descent algorithm, along with its variations such as stochastic gradient descent, is one of the most powerful and popular optimization methods used for deep learning.\n\n## 1.1: Gradient Descent\n\n", "_____no_output_____" ] ], [ [ "#@title Video 1.1: Introduction\nfrom IPython.display import YouTubeVideo\nvideo = YouTubeVideo(id=\"PFQeUDxQFls\", width=854, height=480, fs=1)\nprint(\"Video available at https://youtu.be/\" + video.id)\nvideo", "_____no_output_____" ], [ "#@title Video 1.2: Gradient Descent\nfrom IPython.display import YouTubeVideo\nvideo = YouTubeVideo(id=\"Z3dyRLR8GbM\", width=854, height=480, fs=1)\nprint(\"Video available at https://youtu.be/\" + video.id)\nvideo", "_____no_output_____" ] ], [ [ "\nGradient Descent (introduced by Augustin-Louis Cauchy in 1847) is an **iterative method** to **minimize** a **continuous** and (ideally) **differentiable function** of **many variables**.\n\n### Definition\nLet $f(\\mathbf{w}): \\mathbb{R}^d \\rightarrow \\mathbb{R}$ be a differentiable function. Gradient Descent is an iterative algorithm for minimizing the function $f$, starting with an initial value for variables $\\mathbf{w}$, taking steps of size $\\eta$ in the direction of the negative gradient at the current point to update the variables $\\mathbf{w}$.\n\n$$ \\mathbf{w}^{(t+1)} = \\mathbf{w}^{(t)} - \\eta \\nabla f (\\mathbf{w}^{(t)}) $$\n\nwhere $\\eta > 0$ and $\\nabla f (\\mathbf{w})= \\left( \\frac{\\delta f(\\mathbf{w})}{\\delta w_1}, ..., \\frac{\\delta f(\\mathbf{w})}{\\delta w_d} \\right)$. Since negative gradients always point locally in the direction of steepest descent, the algorithm makes small steps at each point **towards** the minimum.\n \n<br/>\n\n### Vanilla Algorithm\n\n---\n> *inputs*: initial guess $\\mathbf{w}^{(0)}$, step size $\\eta > 0$, number of steps $T$\n\n> *For* *t = 1, 2, ..., T* *do* \\\n$\\qquad$ $\\mathbf{w}^{(t+1)} = \\mathbf{w}^{(t)} - \\eta \\nabla f (\\mathbf{w}^{(t)})$\\\n*end*\n\n> *return*: $\\mathbf{w}^{(t+1)}$\n\n---\n\n<br/>\n\nTo be able to use this algorithm, we need to calculate the gradient of the loss with respect to the learnable parameters.\n", "_____no_output_____" ], [ "## 1.2: Calculating Gradients\n\nTo minimize the empirical risk (loss) function using gradient descent, we need to calculate the vector of partial derivatives:\n\n$$\\dfrac{\\partial Loss}{\\partial \\mathbf{w}} = \\left[ \\dfrac{\\partial Loss}{\\partial w_1}, \\dfrac{\\partial Loss}{\\partial w_2} , ..., \\dfrac{\\partial Loss}{\\partial w_d} \\right]^{\\top} $$\n\nAlthough PyTorch and other deep learning frameworks (e.g. JAX and TensorFlow) provide us with incredibly powerful and efficient algorithms for automatic differentiation, calculating few derivatives with hand would be fun.\n\n### Exercise 1.2\n1. Given $L(w_1, w_2) = w_1^2 - 2w_1 w_2$ find $\\dfrac{\\partial L}{\\partial w_1}$ and $\\dfrac{\\partial L}{\\partial w_1}$.\n\n<br/>\n\n2. Given $f(x) = sin(x)$ and $g(x) = \\ln(x)$, find the derivative of their composite function $\\dfrac{d (f \\circ g)(x)}{d x}$ (*hint: chain rule*).\n\n **Chain rule**: For a composite function $F(x) = f(g(x)) \\equiv (f \\circ g)(x)$:\n$$F'(x) = f'(g(x)) \\cdot g'(x)$$\nor differently denoted:\n$$ \\frac{dF}{dx} = \\frac{df}{dg} ~ \\frac{dg}{dx} $$\n\n<br/>\n\n3. Given $f(x, y, z) = \\tanh \\left( \\ln \\left[1 + z \\frac{2x}{sin(y)} \\right] \\right)$, how easy is it to derive $\\dfrac{\\partial f}{\\partial x}$, $\\dfrac{\\partial f}{\\partial y}$ and $\\dfrac{\\partial f}{\\partial z}$? (*hint: you don't have to actually calculate them!*)\n", "_____no_output_____" ], [ "## 1.3: Computational Graphs and Backprop\n", "_____no_output_____" ] ], [ [ "#@title Video 1.3: Computational Graph\nfrom IPython.display import YouTubeVideo\nvideo = YouTubeVideo(id=\"7c8iCHcVgVs\", width=854, height=480, fs=1)\nprint(\"Video available at https://youtu.be/\" + video.id)\nvideo", "_____no_output_____" ] ], [ [ "\nThe last function in *Exercise 1.2* is an example of how overwhelming the derivation of gradients can get, as the number of variables and nested functions increases. This function is still extremely simple compared to the loss functions of modern neural networks. So how do PyTorch and similar frameworks approach such beasts?\n\n### 1.3.1: Computational Graphs\n\nLet’s look at the function again:\n\n$$f(x, y, z) = \\tanh \\left(\\ln \\left[1 + z \\frac{2x}{sin(y)} \\right] \\right)$$\n\nwe can build a so-called computational graph (shown below) to break the original function to smaller and more approachable expressions. If this \"reverse engineering\" approach seems unintuitive and arbitrary, it's because it is! Usually, the graph is built first.\n\n<center><img src=\"https://raw.githubusercontent.com/ssnio/statics/main/neuromatch/comput_graph.png\" alt=\"Computation Graph\" width=\"800\"/></center>\n\nStarting from $x$, $y$, and $z$ and following the arrows and expressions, you would see that our graph returns the same function as $f$. It does so by calculating intermediate variables $a,b,c,d,$ and $e$. This is called the **forward pass**.\n\nNow, let’s start from $f$, and work our way against the arrows while calculating the gradient of each expression as we go. This is called **backward pass**, which later inspires **backpropagation of errors**.\n\n<center><img src=\"https://raw.githubusercontent.com/ssnio/statics/main/neuromatch/comput_graph_full.png\" alt=\"Computation Graph full\" width=\"1200\"/></center>\n\nBecause we've split the computation into simple operations on intermediate variables, I hope you can appreciate how easy it now is to calculate the partial derivatives. \n\nNow we can use chain rule and simply calculate any gradient:\n\n$$ \\dfrac{\\partial f}{\\partial x} = \\dfrac{\\partial f}{\\partial e}~\\dfrac{\\partial e}{\\partial d}~\\dfrac{\\partial d}{\\partial c}~\\dfrac{\\partial c}{\\partial a}~\\dfrac{\\partial a}{\\partial x} = \\left( 1-\\tanh^2(e) \\right) \\cdot \\frac{1}{d}\\cdot z \\cdot \\frac{1}{b} \\cdot 2$$\n\nConveniently, the values for $e$, $b$, and $d$ are available to us from when we did the forward pass through the graph. That is, the partial derivatives have simple expressions in terms of the intermediate variables $a,b,c,d,e$ that we calculated and stored during the forward pass.\n", "_____no_output_____" ], [ "### Exercise 1.3\nFor the function above, calculate the $\\dfrac{\\partial f}{\\partial y}$ and $\\dfrac{\\partial f}{\\partial z}$ using the computational graph and chain rule.", "_____no_output_____" ], [ "For more: [Calculus on Computational Graphs: Backpropagation](https://colah.github.io/posts/2015-08-Backprop/)", "_____no_output_____" ], [ "---\n# Section 2: PyTorch AutoGrad", "_____no_output_____" ] ], [ [ "#@title Video 2.1: Automatic Differentiation\nfrom IPython.display import YouTubeVideo\nvideo = YouTubeVideo(id=\"h8B8Nlcz7yY\", width=854, height=480, fs=1)\nprint(\"Video available at https://youtu.be/\" + video.id)\nvideo", "_____no_output_____" ] ], [ [ "Deep learning frameworks such as PyTorch, JAX, and TensorFlow come with a very efficient and sophisticated set of algorithms, commonly known as Automatic differentiation. AutoGrad is PyTorch's automatic differentiation engine. Here we start by covering the essentials of AutoGrad, but you will learn more in the coming days.\n\n", "_____no_output_____" ], [ "## Section 2.1: Forward Propagation\n\nEverything starts with the forward propagation (pass). PyTorch plans the computational graph, as we declare the variables and operations, and it builds the graph when we call the backward pass. PyTorch rebuilds the graph every time we iterate or change it (or simply put, PyTorch uses a dynamic graph).\n\nBefore we start our first example, let's recall gradient descent algorithm. In gradient descent algorithm, it is only required to have the gradient of our cost function with respect to variables which are accessible to us for updating (changing). These variables are often called \"learnable parameters\" or simply parameter in PyTorch. In the case of neural networks, weights and biases are often the learnable parameters.", "_____no_output_____" ], [ "### Exercise 2.1\n\nIn PyTorch, we can set the optional argument `requires_grad` in tensors to `True`, so PyTorch would track every operation on them while configuring the computational graph. For this exercise, use the provided tensors to build the following graph.\n\n<br/>\n\n<center><img src=\"https://raw.githubusercontent.com/ssnio/statics/main/neuromatch/simple_graph.png\" alt=\"Simple nn graph\" width=\"600\"/></center>", "_____no_output_____" ] ], [ [ "class SimpleGraph:\n def __init__(self, w=None, b=None):\n \"\"\"Initializing the SimpleGraph\n\n Args:\n w (float): initial value for weight\n b (float): initial value for bias\n \"\"\"\n if w is None:\n self.w = torch.randn(1, requires_grad=True)\n else:\n self.w = torch.tensor([w], requires_grad=True)\n if b is None:\n self.b = torch.randn(1, requires_grad=True)\n else:\n self.b = torch.tensor([b], requires_grad=True)\n\n def forward(self, x):\n \"\"\"Forward pass\n\n Args:\n x (torch.Tensor): 1D tensor of features\n\n Returns:\n torch.Tensor: model predictions\n \"\"\"\n assert isinstance(x, torch.Tensor)\n #################################################\n ## Implement the the forward pass to calculate prediction\n ## Note that prediction is not the loss, but the value after `tanh`\n # Complete the function and remove or comment the line below\n raise NotImplementedError(\"Forward Pass `forward`\")\n #################################################\n prediction = ...\n return prediction\n\n\ndef sq_loss(y_true, y_prediction):\n \"\"\"L2 loss function\n\n Args:\n y_true (torch.Tensor): 1D tensor of target labels\n y_true (torch.Tensor): 1D tensor of predictions\n\n Returns:\n torch.Tensor: L2-loss (squared error)\n \"\"\"\n assert isinstance(y_true, torch.Tensor)\n assert isinstance(y_prediction, torch.Tensor)\n #################################################\n ## Implement the L2-loss (squred error) given true label and prediction\n # Complete the function and remove or comment the line below\n raise NotImplementedError(\"Loss function `sq_loss`\")\n #################################################\n loss = ...\n return loss\n\n\n# # Uncomment to run\n# feature = torch.tensor([1]) # input tensor\n# target = torch.tensor([7]) # target tensor\n\n# simple_graph = SimpleGraph(-0.5, 0.5)\n# print(\"initial weight = {} \\ninitial bias = {}\".format(simple_graph.w.item(),\n# simple_graph.b.item()))\n\n# prediction = simple_graph.forward(feature)\n# square_loss = sq_loss(target, prediction)\n\n# print(\"for x={} and y={}, prediction={} and L2 Loss = {}\".format(feature.item(),\n# target.item(),\n# prediction.item(),\n# square_loss.item()))", "_____no_output_____" ] ], [ [ "[*Click for solution*](https://github.com/NeuromatchAcademy/course-content-dl/tree/main//tutorials/W1D2_LinearDeepLearning/solutions/W1D2_Tutorial1_Solution_b9fccdbe.py)\n\n", "_____no_output_____" ], [ "It is important to appreciate the fact that PyTorch can follow our operations as we arbitrary go through classes and functions.", "_____no_output_____" ], [ "## 2.2 Backward Propagation\n\nHere is where all the magic lies. We can first look at the operations that PyTorch kept track of. Tensor property `grad_fn` keeps reference to backward propagation functions.", "_____no_output_____" ] ], [ [ "print('Gradient function for prediction =', prediction.grad_fn)\nprint('Gradient function for loss =', square_loss.grad_fn)", "_____no_output_____" ] ], [ [ "Now let's kick off backward pass to calculate the gradients by calling the `.backward()` on the tensor we wish to initiate the backpropagation from. Often, `.backward()` is called on the loss, which is the last on the graph. Before doing that, let's calculate the loss gradients:\n\n$$\\frac{\\partial{loss}}{\\partial{w}} = - 2 x (y_t - y_p)(1 - y_p^2)$$\n\n$$\\frac{\\partial{loss}}{\\partial{b}} = - 2 (y_t - y_p)(1 - y_p^2)$$\n\nWe can then compare it to PyTorch gradients, which can be obtained by calling `.grad` on tensors.\n\n**Important Notes**\n* Always keep in mind that PyTorch is tracking all the operations for tensors that require grad. To stop this tracking, we use `.detach()`.\n\n* PyTorch builds the graph only when `.backward()` is called and then it is set free. If you try calling `.backward()` after it is already called, you get the following error:\n\n *`Trying to backward through the graph a second time, but the saved intermediate results have already been freed. Specify retain_graph=True when calling .backward() or autograd.grad() the first time.`*\n\n* Learnable parameters are \"contagious\". If you recall from our computational graph, we need all the intermediate gradients to be able to use the chain rule. Therefore, we need to `.detach()` any tensor that was on the path of gradient flow (e.g. prediction tensor).\n\n* `.backward()` accumulates gradients in the leaves. For most of training methods, we call `.zero_grad()` on the loss or optimizer to zero `.grad` attributes (see [autograd.backward](https://pytorch.org/docs/stable/autograd.html#torch.autograd.backward) for more).", "_____no_output_____" ] ], [ [ "# analytical gradients (remember detaching)\nana_dloss_dw = - 2 * feature * (target - prediction.detach())*(1 - prediction.detach()**2)\nana_dloss_db = - 2 * (target - prediction.detach())*(1 - prediction.detach()**2)\n\nsquare_loss.backward() # first we should call the backward to build the graph\nautograd_dloss_dw = simple_graph.w.grad\nautograd_dloss_db = simple_graph.b.grad\n\nprint(ana_dloss_dw == autograd_dloss_dw)\nprint(ana_dloss_db == autograd_dloss_db)", "_____no_output_____" ] ], [ [ "References and more:\n* [A GENTLE INTRODUCTION TO TORCH.AUTOGRAD](https://pytorch.org/tutorials/beginner/blitz/autograd_tutorial.html)\n\n* [AUTOMATIC DIFFERENTIATION PACKAGE - TORCH.AUTOGRAD](https://pytorch.org/docs/stable/autograd.html)\n\n* [AUTOGRAD MECHANICS](https://pytorch.org/docs/stable/notes/autograd.html)\n\n* [AUTOMATIC DIFFERENTIATION WITH TORCH.AUTOGRAD](https://pytorch.org/tutorials/beginner/basics/autogradqs_tutorial.html)", "_____no_output_____" ], [ "---\n# Section 3: PyTorch's Neural Net module (`nn.Module`)", "_____no_output_____" ] ], [ [ "#@title Video 3.1: Putting it together\nfrom IPython.display import YouTubeVideo\nvideo = YouTubeVideo(id=\"rUChBWj9ihw\", width=854, height=480, fs=1)\nprint(\"Video available at https://youtu.be/\" + video.id)\nvideo", "_____no_output_____" ] ], [ [ "In this section we will focus on training the simple neural network model from yesterday.", "_____no_output_____" ] ], [ [ "#@title Generate the sample dataset\nimport sklearn.datasets\n\n# Create a dataset of 256 points with a little noise\nX_orig, y_orig = sklearn.datasets.make_moons(256, noise=0.1)\n\n# Plot the dataset\nplt.figure(figsize=(9, 7))\nplt.scatter(X_orig[:,0], X_orig[:,1], s=40, c=y_orig)\nplt.xlabel(\"$x_1$\")\nplt.ylabel(\"$x_2$\")\nplt.show()\n\n# Select the appropriate device (GPU or CPU)\ndevice = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n\n# Convert the 2D points to a float tensor\nX = torch.from_numpy(X_orig).type(torch.FloatTensor)\nX = X.to(device)\n\n# Convert the labels to a long interger tensor\ny = torch.from_numpy(y_orig).type(torch.LongTensor)\ny = y.to(device)", "_____no_output_____" ] ], [ [ "Let's define the same simple neural network model as in Day 1. This time we will not define a `train` method, but instead implement it outside of the class so we can better inspect it.", "_____no_output_____" ] ], [ [ "# Simple neural network with a single hidden layer\nclass NaiveNet(nn.Module):\n\n def __init__(self):\n \"\"\"\n Initializing the NaiveNet\n \"\"\"\n super(NaiveNet, self).__init__()\n self.layers = nn.Sequential(\n nn.Linear(2, 16),\n nn.ReLU(),\n nn.Linear(16, 2),\n )\n\n def forward(self, x):\n \"\"\"Forward pass\n\n Args:\n x (torch.Tensor): 2D tensor of features\n\n Returns:\n torch.Tensor: model predictions\n \"\"\"\n return self.layers(x)", "_____no_output_____" ] ], [ [ "PyTorch provides us with ready to use neural network building blocks, such as linear or recurrent layers, activation functions and loss functions, packed in the [`torch.nn`](https://pytorch.org/docs/stable/nn.html) module. If we build a neural network using the `torch.nn` layers, the weights and biases are already in `requires_grad` mode. \n\nNow let's prepare the training! We need 3 things for that:\n\n* **Model parameters** - Model parameters refer to all the learnable parameters' of the model which are accessible by calling `.parameters()` on the model. Please note that not all the `requires_grad` are seen as model parameters. To create a custom model parameter, you can use [`nn.Parameter`](https://pytorch.org/docs/stable/generated/torch.nn.parameter.Parameter.html) (*A kind of Tensor that is to be considered a module parameter*). When we create a new instace of our model, layer parameters are initialized using a uniform distribution (more on that in the coming tutorials and days).\n\n* **Loss function** - we need to define the loss that we are going to be optimizing. The cross entropy loss is suitable for classification problems.\n\n* **Optimizer** - the optimizer will perform the adaptation of the model parameters according to the chosen loss function. The optimizer takes the parameters of the model (often by calling `.parameters()` on the model) as its input to be adapted.\n\nYou will learn more details about choosing the right loss function and optimizer later in the course.", "_____no_output_____" ] ], [ [ "# Create an instance of our network\nnaive_model = NaiveNet().to(device)\n\n# Create a cross entropy loss function\ncross_entropy_loss = nn.CrossEntropyLoss()\n\n# Stochstic Gradient Descent optimizer with a learning rate of 1e-1\nlearning_rate = 1e-1\nsgd_optimizer = torch.optim.SGD(naive_model.parameters(), lr=learning_rate)", "_____no_output_____" ] ], [ [ "The training process in PyTorch is interactive - you can perform training iterations as you wish and inspect the results after each iteration. We encourage leaving the loss function outside the explicit forward pass function, and rather calculate it on the output (prediction).\n\nLet's perform one training iteration. You can run the cell multiple times and see how the parameters are being updated and the loss is reducing. We pick the parameters of the first neuron in the first layer. Please make sure you go through all the commands and discuss their purpose with the pod.", "_____no_output_____" ] ], [ [ "# Reset all gradients to 0\nsgd_optimizer.zero_grad()\n\n# Forward pass (Compute the output of the model on the data)\ny_logits = naive_model(X)\n\n# Compute the loss\nloss = cross_entropy_loss(y_logits, y)\nprint('Loss:', loss.item())\n\n# Perform backpropagation to build the graph and compute the gradients\nloss.backward()\n\n# `.parameters()` returns a generator\nprint('Gradients:', next(naive_model.parameters()).grad[0].detach().numpy())\n\n# Print model's first learnable parameters\nprint('Parameters before:', next(naive_model.parameters())[0].detach().numpy())\n\n# Optimizer takes a step in the steepest direction (negative of gradient)\n# and \"updates\" the weights and biases of the network\nsgd_optimizer.step()\n\n# Print model's first learnable parameters\nprint('Parameters after: ', next(naive_model.parameters())[0].detach().numpy())", "_____no_output_____" ] ], [ [ "## Exercise 3\nFollowing everything we learned so far, we ask you to complete the `train` function below.", "_____no_output_____" ] ], [ [ "def train(features, labels, model, loss_fun, optimizer, n_epochs):\n\n \"\"\"Training function\n\n Args:\n features (torch.Tensor): features (input) with shape torch.Size([n_samples, 2])\n labels (torch.Tensor): labels (targets) with shape torch.Size([n_samples, 2])\n model (torch nn.Module): the neural network\n loss_fun (function): loss function\n optimizer(function): optimizer\n n_epochs (int): number of training epochs\n\n Returns:\n list: record (evolution) of losses\n list: record (evolution) of value of the first parameter\n list: record (evolution) of gradient of the first parameter\n \"\"\"\n loss_record = [] # keeping recods of loss\n par_values = [] # keeping recods of first parameter\n par_grads = [] # keeping recods of gradient of first parameter\n\n # we use `tqdm` methods for progress bar\n epoch_range = trange(n_epochs, desc='loss: ', leave=True)\n for i in epoch_range:\n\n if loss_record:\n epoch_range.set_description(\"loss: {:.4f}\".format(loss_record[-1]))\n epoch_range.refresh() # to show immediately the update\n time.sleep(0.01)\n\n #################################################\n ## Implement the missing parts of the training loop\n # Complete the function and remove or comment the line below\n raise NotImplementedError(\"Training setup `train`\")\n #################################################\n ... # Initialize gradients to 0\n predictions = ... # Compute model prediction (output)\n loss = ... # Compute the loss\n ... # Compute gradients (backward pass)\n ... # update parameters (optimizer takes a step)\n\n loss_record.append(loss.item())\n par_values.append(next(model.parameters())[0][0].item())\n par_grads.append(next(model.parameters()).grad[0][0].item())\n\n return loss_record, par_values, par_grads\n\n# # Uncomment to run\n# epochs = 5000\n# losses, values, gradients = train(X, y,\n# naive_model,\n# cross_entropy_loss,\n# sgd_optimizer,\n# epochs)\n\n# ex3_plot(epochs, losses, values, gradients)", "_____no_output_____" ] ], [ [ "[*Click for solution*](https://github.com/NeuromatchAcademy/course-content-dl/tree/main//tutorials/W1D2_LinearDeepLearning/solutions/W1D2_Tutorial1_Solution_364cd4e2.py)\n\n*Example output:*\n\n<img alt='Solution hint' align='left' width=704 height=488 src=https://raw.githubusercontent.com/NeuromatchAcademy/course-content-dl/main/tutorials/W1D2_LinearDeepLearning/static/W1D2_Tutorial1_Solution_364cd4e2_2.png>\n\n", "_____no_output_____" ] ], [ [ "#@title Video 3.2: Wrap-up\nfrom IPython.display import YouTubeVideo\nvideo = YouTubeVideo(id=\"zFmWs6doqhM\", width=854, height=480, fs=1)\nprint(\"Video available at https://youtu.be/\" + video.id)\nvideo", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "## Classifying Hourly Timestamps into Day or Night", "_____no_output_____" ] ], [ [ "import pandas as pd\nimport numpy as np\nimport warnings\nfrom df2gspread import df2gspread as d2g\nwarnings.simplefilter('ignore') ", "_____no_output_____" ], [ "## Import Raw TTN Data from Google SpreadSheet\nurl = 'https://docs.google.com/spreadsheets/d/e/2PACX-1vRlXVQ6c3fKWvtQlFRSRUs5TI3soU7EghlypcptOM8paKXcUH8HjYv90VoJBncuEKYIZGLq477xE58C/pub?gid=0&single=true&output=csv'\ndf_hourly = pd.read_csv(url,parse_dates = ['time'],infer_datetime_format = True,usecols = [0,3])\ndf_hourly.head()", "_____no_output_____" ], [ "## Cleaning and re-organizing the DataFrame\ndf_hourly.rename(columns={'time': 'TimeStamps'}, inplace=True)\ndf_hourly.rename(columns={'CarCount': 'VehicleCountperHour'}, inplace=True)\n\n## Strip the Microseconds from the time column\ndf_hourly['TimeStamps'] = df_hourly['TimeStamps'].values.astype('datetime64[s]')\n\n## Reorder the dataframe\ndf_hourly = df_hourly.reindex(['TimeStamps','VehicleCountperHour'], axis=1)\ndf_hourly.tail()", "_____no_output_____" ] ], [ [ "### Importing SunTime Chart for adding Day or Night Classification\n Here we add Day or Night classification to Hourly Vehicle Count Data using sunrise and sunset times dataframe of Saarbrucken. These times were acquired from [timeanddate.com](https://www.timeanddate.com/sun/germany/saarbrucken) and the dataframe was manually created.", "_____no_output_____" ] ], [ [ "# Reading SunTime Chart of Saarbrücken\nurl = 'https://docs.google.com/spreadsheets/d/e/2PACX-1vRJCZQkgUHTCcx-wvJOog87qq4EFiQ1W6T4akxLSpiqCb3KjYDf_43coltDGG0YcjjsDTxjeXE-O_NH/pub?gid=0&single=true&output=csv'\ndf_suntimes = pd.read_csv(url)\n# Modified Sun Timings DataFrame\ndf_suntimes_mod = pd.DataFrame()\ndf_suntimes_mod['SunriseTimeStamp']= pd.to_datetime(df_suntimes['Date'] + ' ' + df_suntimes['Sunrise'])\ndf_suntimes_mod['SunsetTimeStamp']= pd.to_datetime(df_suntimes['Date'] + ' ' + df_suntimes['Sunset'])\n# Querying values in the dataframe to selected dates\nstart_time = '2018-02-21 07:25:00'\ndf_suntimes_mod = df_suntimes_mod.loc[df_suntimes_mod.SunriseTimeStamp >= start_time,:]\nend_time = '2018-02-25 18:15:00'\ndf_suntimes_mod = df_suntimes_mod.loc[df_suntimes_mod.SunriseTimeStamp <= end_time,:]\ndf_suntimes_mod = df_suntimes_mod.reset_index()\ndf_suntimes_mod = df_suntimes_mod.drop(['index'], 1)\ndf_suntimes_mod", "_____no_output_____" ], [ "## Creating a new Dataframe from original dataframe to classify Hourly TimeStamps as 'Day' or 'Night':\ndf_dn = df_hourly\n\n## Set Everything to Day First\ndf_dn['DayorNight'] = 'Day'\n\n## Manually fixing first day's night timestamps to 'Night'\nnight_index = (df_dn.loc[df_dn.TimeStamps <= (df_suntimes_mod['SunriseTimeStamp'][0]),:]).index\n\n# Select Night Time Traffic Only from Sunset today to next day Sunrise\nn_days = len(df_suntimes_mod['SunriseTimeStamp'])\nfor i,j in zip(range(n_days),range(1,n_days)): \n start_time = df_suntimes_mod['SunsetTimeStamp'][i]\n end_time = df_suntimes_mod['SunriseTimeStamp'][j]\n data = df_dn[(df_dn['TimeStamps'] > start_time) & (df_dn['TimeStamps'] < end_time)]\n night_index = night_index.append(data.index)\n \n## Set all the Night TimeStamps to 'Night\ndf_dn['DayorNight'].iloc[night_index] = 'Night'\ndf_dn.head(15)", "_____no_output_____" ], [ "# Writing the file as csv\ndf_dn.to_csv('data/DayorNight.csv', date_format=\"%d/%m/%Y %H:%M:%S\",index=False)", "_____no_output_____" ], [ "## Write pandas dataframe to a Google Sheet Using df2spread:\n\n# Insert ID of Google Spreadsheet\nspreadsheet = '1LTXIPNb7MX0qEOU_DbBKC-OwE080kyRvt-i_ejFM-Yg'\n\n# Insert Sheet Name\nwks_name = 'CleanedData'\n\nd2g.upload(df_dn,spreadsheet,wks_name,col_names=True,clean=True)", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "## Coding Matrices\n\nHere are a few exercises to get you started with coding matrices. The exercises start off with vectors and then get more challenging\n\n### Vectors", "_____no_output_____" ] ], [ [ "### TODO: Assign the vector <5, 10, 2, 6, 1> to the variable v\nv = []", "_____no_output_____" ] ], [ [ "The v variable contains a Python list. This list could also be thought of as a 1x5 matrix with 1 row and 5 columns. How would you represent this list as a matrix?", "_____no_output_____" ] ], [ [ "### TODO: Assign the vector <5, 10, 2, 6, 1> to the variable mv\n### The difference between a vector and a matrix in Python is that\n### a matrix is a list of lists.\n\n### Hint: See the last quiz on the previous page\n\nmv = [[]]", "_____no_output_____" ] ], [ [ "How would you represent this vector in its vertical form with 5 rows and 1 column? When defining matrices in Python, each row is a list. So in this case, you have 5 rows and thus will need 5 lists.\n\nAs an example, this is what the vector $$<5, 7>$$ would look like as a 1x2 matrix in Python: \n```python\nmatrix1by2 = [\n [5, 7]\n]\n```\n\nAnd here is what the same vector would look like as a 2x1 matrix:\n```python\nmatrix2by1 = [\n [5], \n [7]\n]\n```", "_____no_output_____" ] ], [ [ "### TODO: Assign the vector <5, 10, 2, 6, 1> to the variable vT\n### vT is a 5x1 matrix\nvT = []", "_____no_output_____" ] ], [ [ "### Assigning Matrices to Variables", "_____no_output_____" ] ], [ [ "### TODO: Assign the following matrix to the variable m\n### 8 7 1 2 3\n### 1 5 2 9 0\n### 8 2 2 4 1\n\nm = [[]]", "_____no_output_____" ] ], [ [ "### Accessing Matrix Values", "_____no_output_____" ] ], [ [ "### TODO: In matrix m, change the value \n### in the second row last column from 0 to 5\n### Hint: You do not need to rewrite the entire matrix\n", "_____no_output_____" ] ], [ [ "### Looping through Matrices to do Math\n\nCoding mathematical operations with matrices can be tricky. Because matrices are lists of lists, you will need to use a for loop inside another for loop. The outside for loop iterates over the rows and the inside for loop iterates over the columns.\n\n\nHere is some pseudo code\n```python\nfor i in number of rows:\n for j in number of columns:\n mymatrix[i][j]\n```\n\nTo figure out how many times to loop over the matrix, you need to know the number of rows and number of columns. \n\n\nIf you have a variable with a matrix in it, how could you figure out the number of rows? How could you figure out the number of columns? The [len](https://docs.python.org/2/library/functions.html#len) function in Python might be helpful.", "_____no_output_____" ], [ "### Scalar Multiplication", "_____no_output_____" ] ], [ [ "### TODO: Use for loops to multiply each matrix element by 5\n### Store the answer in the r variable. This is called scalar\n### multiplication\n###\n### HINT: First write a for loop that iterates through the rows\n### one row at a time\n###\n### Then write another for loop within the for loop that\n### iterates through the columns\n###\n### If you used the variable i to represent rows and j\n### to represent columns, then m[i][j] would give you\n### access to each element in the matrix\n###\n### Because r is an empty list, you cannot directly assign\n### a value like r[i][j] = m[i][j]. You might have to\n### work on one row at a time and then use r.append(row).\nr = []", "_____no_output_____" ] ], [ [ "### Printing Out a Matrix", "_____no_output_____" ] ], [ [ "### TODO: Write a function called matrix_print() \n### that prints out a matrix in\n### a way that is easy to read.\n### Each element in a row should be separated by a tab\n### And each row should have its own line\n### You can test our your results with the m matrix\n\n### HINT: You can use a for loop within a for loop\n### In Python, the print() function will be useful\n### print(5, '\\t', end = '') will print out the integer 5, \n### then add a tab after the 5. The end = '' makes sure that\n### the print function does not print out a new line if you do\n### not want a new line.\n\n### Your output should look like this\n### 8 7 1 2 3 \n### 1 5 2 9 5 \n### 8 2 2 4 1\n\ndef matrix_print(matrix):\n return\n\nm = [\n [8, 7, 1, 2, 3],\n [1, 5, 2, 9, 5],\n [8, 2, 2, 4, 1]\n]\n\nmatrix_print(m)", "_____no_output_____" ] ], [ [ "### Test Your Results", "_____no_output_____" ] ], [ [ "### You can run these tests to see if you have the expected\n### results. If everything is correct, this cell has no output\n\nassert v == [5, 10, 2, 6, 1]\nassert mv == [\n [5, 10, 2, 6, 1]\n]\n\nassert vT == [\n [5], \n [10], \n [2], \n [6], \n [1]]\n\nassert m == [\n [8, 7, 1, 2, 3], \n [1, 5, 2, 9, 5], \n [8, 2, 2, 4, 1]\n]\n\nassert r == [\n [40, 35, 5, 10, 15], \n [5, 25, 10, 45, 25], \n [40, 10, 10, 20, 5]\n]", "_____no_output_____" ] ], [ [ "### Print Out Your Results", "_____no_output_____" ] ], [ [ "### Run this cell to print out your answers\nprint(v)\nprint(mv)\nprint(vT)\nprint(m)\nprint(r)", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "%load_ext sql\n%sql sqlite:///flights.db", "_____no_output_____" ] ], [ [ "숙제 1\n=======\n\n### 일러두기 :\n\n**_꼼꼼하게 읽어보기 바랍니다_**\n\n\n* `prettytable` 모듈을 설치해야 스크립트를 실행할 수 있음. (설치 방법: `pip install --user prettytable`)\n * `flights.db` 파일이 숙제용 Jupyter notebook과 같은 디렉터리에 있어야 함 (없다면 [여기서](http://open.gnu.ac.kr/lecslides/2018-2-DB/Assignments1/flights.db.zip) 다운 받기) 압축을 해제해야 함. `flights.db.zip`이 있는 곳에서 `unzip flights.db.zip`으로 압축을 해제하면 됨\n* 데이터베이스 `flights.db`를 다운 받은 후 가장 위의 셀의 명령 실행하기\n* 테스트, 디버그, 탐색하기 등을 위해서 새로운 셀을 생성하는 것을 적극 권장함\n* 셀을 실행시키고 셀 왼 편에 `In [*]:` 이 보인다면 _실행 중_ 을 의미함\n * **만약 셀이 오랫 동안 결과를 내 놓지 않고 멈춘 것 같다면: SQL 에 다시 연결하도록 python kernel을 다시 시작해야 함**\n * 커널을 다시 시작하는 방법: \"Kernel >> Restart & Clear Output\", 그리고 위의 셀부터 아래로 하나씩 실행 함 \n * 다른 버전의 데이터베이스를 로드하기 위해서도 마찬가지를 새로운 연결을 만들어야 함\n* 기억하기:\n * `%sql [SQL 질의문;]` 은 _한 줄짜리_ SQL 질의문에 사용\n * `%%sql \n [SQL 질의문;]` 은 _여러 줄짜리_ SQL 질의문에 사용\n* `submit.py` 을 실행하면 질의문을 처리하고 출력함 \n* 실행의 결과는 `correct_output.txt` 파일에 나와 있음.\n * 실행 결과의 비교를 원한다면 `python sanity_check.py` 을 실행하거나, 다음의 명령을 실행하여 결과를 얻을 수 있음 `python submit.py > my_output; diff my_output correct_output.txt` 터미널에서 입력해야 함\n * **숙제로 작성한 `submit.py` 파일은 아래의 형식을 절대적으로 따라야 함.** 형식이라 함은:\n * 컬럼의 이름은 `correct_output.txt` 에 나와 있는 이름과 **똑같은 이름**이어야 함\n * 컬럼의 순서도 `correct_output.txt` 에 나와 있는 순서와 **똑같은 순서**이어야 함 \n\n### 제출 방법:\n * iPython notebook 자체를 제출하지 말 것\n * 대신에, `submit.py` 에 작성한 번호에 맞게 질의문을 복사 붙여 넣기 할 것\n * `%sql` 또는 `%%sql` 명령은 SQL 문에 포함시키지 말 것\n * 제출한 질의문을 똑같은 스키마에서 임의로 선택된 값에 대상으로 실행시켜 평가를 할 것임. 그렇기 때문에 해답과 똑같은 결과가 나오도록 상수등을 써서 조작하지 말것\n * **`submission_instructions.txt` 에 설명된 방법으로 해답을 제출할 것**\n\n_즐겁게 시작해봅시다!_", "_____no_output_____" ], [ "개요: 여행 일정 지연\n------------------------\n\n여행 일정이 지연 되는 것만큼 짜증 나는 일은 없습니다. 그렇지 않나요?\n\n여행 일정이 지연되지 않도록 여러가지 새로운 방법을 찾아봅니다. 최근에 찾은 데이터가 왜 일정이 지연되는지 이유와 무엇을 포기할지를 잘 설명해주고 있습니다. \n\nSQL을 사용해서 한 번 그 이유들을 알아봅시다.", "_____no_output_____" ], [ "----", "_____no_output_____" ], [ "이 과제에서는 2017년 7월의 여객기의 지연 정보의 정보를 담고 있습니다. 데이터베이스의 기본 릴레이션에 대한 정보를 알아 봅시다.", "_____no_output_____" ] ], [ [ "%%sql\nSELECT * \nFROM flight_delays \nLIMIT 1;", " * sqlite:///flights.db\nDone.\n" ] ], [ [ "굉장히 많은 컬럼들이 있는 것을 알 수 있는데, 그러면 몇 줄이나 될까요?", "_____no_output_____" ] ], [ [ "%%sql\nSELECT COUNT(*) AS num_rows\nFROM flight_delays", " * sqlite:///flights.db\nDone.\n" ] ], [ [ "데이터의 양이 상당합니다! 손과 머리로만 해답을 찾지 못할 것 같군요. \n\n데이터베이스에 더 많은 데이터를 넣을 필요는 없겠군요. 컬럼들이 어떤 의미를 갖는지 알아 보려면 [이 링크](https://www.transtats.bts.gov/DL_SelectFields.asp?Table_ID=236)를 따라가기 바랍니다. \n\n몇 개의 추가적인 테이블들을 같이 포함해 놓았습니다. 이 테이블들을 사용하면 `airline_id`, `airport_id`, 그리고 `day_of_week` 을 사람이 읽기 편한 정보로 변환할 수 있습니다. \n\n아래의 셀을 이용하여 `airlines`과 `weekdays` 의 정보를 확인해보기 바랍니다:", "_____no_output_____" ] ], [ [ "%%sql\n", "_____no_output_____" ] ], [ [ "좋습니다. 이제 시작해봅시다.", "_____no_output_____" ], [ "# SQL 질의문", "_____no_output_____" ], [ "질의문 1: 항공편의 평균 지연 시간은? \n------------------------\n데이터에 대한 이해를 돕기 위해, 간단한 질의문을 작성해봅시다.\n\n아래의 셀에 2017년 7월동안 모든 항공편의 평균 지연시간을 구하는 질의문을 작성해봅시다.", "_____no_output_____" ] ], [ [ "%%sql\n", "_____no_output_____" ] ], [ [ "질의문 2: 가장 긴 지연 시간은?\n------------------------\n평균은 그리 크지 않군요. 하지만 _최장_ 지연 시간은 어떻게 되나요?\n\n아래의 셀에 2017년 7월동안 가장 늦게 도착한 시간을 찾는 질의문을 작성해봅시다.", "_____no_output_____" ] ], [ [ "%%sql\n", "_____no_output_____" ] ], [ [ "질의문 3: 어떤 항공편을 피하는 것이 정신 건강에 좋을까요?\n------------------------\n\n어떤 항공편이 가장 늦었나요?\n\n\n아래의 셀에 2017년 7월에 가장 늦게 도착한 항공사(`carrier`)와 항공편 명, 출발 도시 명, 도착 도시 명, 항공 일정을 출력하는 질의문을 작성 바랍니다. 앞에서 얻은 정보를 질의문에 삽입해서 계산하지 말고 중첩 질의문을 쓰기 바랍니다.", "_____no_output_____" ] ], [ [ "%%sql\n", "_____no_output_____" ] ], [ [ "질의문 4: 어떤 요일이 여행하기 가장 안좋은 날인가요?\n------------------------\n\n학기가 시작되었으니 먼 곳으로 여행을 할 수는 없지만, 출장은 가야하겠지요. 비행기를 타기 가장 안좋은 날은 무슨 요일인가요?\n\n아래의 셀에 요일마다 평균 지연 시간이 어떻게 되는지 내림차순으로 정렬하여 결과를 출력하도록 질의문을 작성하기 바랍니다. 출력 결과의 스키마는 (`weekday_name`, `average_delay`)의 형태를 갖고 있어야 합니다.\n\n**Note: 요일의 ID를 그대로 출력하지 말기 바랍니다.** (Hint: `weekdays` 테이블을 사용하여 join하여 요일의 이름을 출력하도록 합시다.)", "_____no_output_____" ] ], [ [ "%%sql\n", "_____no_output_____" ] ], [ [ "질의문 5: SFO에서 출발하는 항공사 중 지연 시간이 가장 긴 항공사는 어디입니까?\n------------------------\n\n어떤 요일을 피해야 할지 알았으니 SFO에서 출발하는 항공사 중 한 곳을 정해야 합니다. 어디로 갈지는 말하지 않았으니, SFO에서 출발하는 모든 항공사의 항공편들의 평균 지연시간을 구해 봅시다.\n\n아래의 셀에 2017년 7월에 SFO에서 출발한 각 항공사 별로 모든 항공편에 대해 평균 지연시간을 내림차순으로 구하는 질의문을 작성해봅시다.\n\n**Note: 항공사 ID를 그대로 출력하지 맙시다.** (Hint: 중첩 질의문으로 `airlines` 테이블을 join 하여 항공사 이름을 출력합시다.)", "_____no_output_____" ] ], [ [ "%%sql\n", "_____no_output_____" ] ], [ [ "질의문 6: 항공사들의 지연 비율을 알아 봅시다\n------------------------\n지연되는 항공편이 많습니다. 어떤 항공사가 지연시간이 많은 알아봅시다.\n\n아래의 셀에 평균 10분 이상 지연되는 항공편이 있었던 항공사들의 비율을 구해봅시다. 전체 항공사의 수를 세서 질의문에 포함시키지 말기 바랍니다. 그리고 질의문에는 최소한 하나 이상의 `HAVING` 절을 사용합시다.\n\nNote: sqlite 의 `COUNT(*)`는 정수형을 리턴하기 때문에 실수형으로 결과를 출력하려면 최소한 한 번 이상 `SELECT CAST (COUNT(*) AS float)` 또는 `COUNT(*)*1.0` 절을 써야 합니다.", "_____no_output_____" ] ], [ [ "%%sql\n", "_____no_output_____" ] ], [ [ "질의문 7: 출발 지연이 도착 지연에 어떤 영향을 미치나요?\n------------------------\n\n비행기가 지연 출발하면 도착 시간에 얼마나 영향을 주는지 알고 싶습니다.\n\n\n[샘플 공분산](https://en.wikipedia.org/wiki/Covariance) 은 두 변수 간의 분산량을 측정하여 상관관계가 있는지 알려주는 통계치입니다. 공분산이 클수록 상관관계가 높고 음수인 경우 역상관관계가 있습니다. 샘플 공분산의 계산 식은 다음과 같습니다:\n$$\nCov(X,Y) = \\frac{1}{n-1} \\sum_{i=1}^n (x_i-\\hat{x})(y_i-\\hat{y})\n$$\n이 때, $x_i$ 는 $X$의 $i$번째 값이고, $y_i$는 $Y$의 $i$번째 값입니다. $X$ 와 $Y$의 평균은 $\\bar{x}$ 과 $\\bar{y}$으로 표현 하였습니다.\n\n아래의 셀에 도착 지연과 출발 지연 시간의 공분산을 구하는 하나의 질의문을 작성 해보기 바랍니다.\n\n*Note: [Pearson correlation coefficient](https://en.wikipedia.org/wiki/Pearson_correlation_coefficient) 으로 구할 수도 있습니다. 그 결과는 정규화 되어 1 부터 -1의 값으로 상관관계를 알려 줍니다. 하지만, SQLite는 루트 계산 함수가 들어 있지 않기 때문에 이 계산식을 쓸 수 가 없습니다. 다른 보편적인 데이터베이스(PostgreSQL와 MySQL)에는 루트 계산 함수가 구현되어 있습니다.*", "_____no_output_____" ] ], [ [ "%%sql\n", "_____no_output_____" ] ], [ [ "질의문 8: 한 주가 엉망이었습니다...\n------------------------\n\n7월 어떤 항공사의 마지막 한 주(24일 이후)의 평균 지연 시간이 그 이전 주(24일 이전)들의 평균 지연 시간보다 절대적으로 길었나요?\n\n아래의 셀에 1일부터 23일까지의 평균 지연 시간 대비 24일 부터 31일 사이의 평균 지연 시간이 절대적으로 길었던 항공사의 이름을 출력하는 질의문을 작성하기 바랍니다.\n\nNote: [sqlite에서 날짜 다루기](http://www.sqlite.org/lang_datefunc.html)에 따라 `day_of_month`을 사용하여 질의문을 작성하는 것이 편할 것입니다.\n\nNote 2: 아마 과제 중 가장 어려운 질의문이 될 수도 있는데, 작은 단위로 질의문을 작성하여 한 부분씩 해결하고, 그 질의문을 합쳐서 최종 질의문을 작성하는 것이 좋습니다.\n\nHint: 두 개의 하위 질의문으로 계산할 수 있습니다. 하나의 질의문이 24일 이후의 평균 도착 시간을 계산하고, 다른 질의문이 24일 이전의 도착 시간을 계산하고, 두 질의문을 join하여 지연 시간의 차를 계산하면 됩니다.", "_____no_output_____" ] ], [ [ "%%sql\n", "_____no_output_____" ] ], [ [ "질의문 9: 진보적인 그리고 혁명적인\n------------------------\n포트랜드 (PDX)와 유진 (EUG)로 가기를 원하지만, 한 번에 가기가 쉽지 않군요. 우수 고객 마일리지를 채우기 위해 같은 항공편으로 각 도시로 이동하기를 원합니다. SFO -> PDF와 SFO -> EUG 로 가는 같은 항공사가 있는지 알고 싶습니다.\n\n아래의 셀에 2017년 7월에 SFO -> PDX 과 SFO -> EUG 을 출항한 항공사의 유일한 이름(중복 없음 ID 가 아님)을 출력하는 하나의 SQL 질의문을 작성하기 바랍니다.", "_____no_output_____" ] ], [ [ "%%sql\n", "_____no_output_____" ] ], [ [ "질의문 10: 피로도와 등거리 간의 결정\n------------------------\n\n시카고에서 캘리포니아로 이동하려고 합니다. Midway (MDW) 또는 O'Hare (ORD) 에서 샌프란시스코 (SFO), 산호세 (SJC), 오크랜드 (OAK)로 도착하면 좋겟습니다. 만약 이 번 달이 7월이라고 하면 시카고에서 현지 시간 14시에 출발하는 경로 중 지연 시간이 가장 짧은 경로가 어떤 것입니까?\n\n아래의 셀에 MDW 또는 ORD 에서 현지 시간 14시(`crs_dep_time`)에 출발하고 SFO, SJC, 또는 OAK에 도착하는 항공편들의 평균 지연 시간을 구하는 하나의 질의문을 작성하기 바랍니다. 출발과 도착 공항을 Group by로 묶고 지연 시간의 내림차순으로 출력하기 바랍니다.\n\nNote: `crs_dep_time` 필드는 정수 형을 갖고 있으며 hhmm (e.g. 4:15pm 은 1615 임) 형을 따름.", "_____no_output_____" ] ], [ [ "%%sql\n", "_____no_output_____" ] ], [ [ "## 다 끝났습니다. 이제 제출합시다.\n * 제출하는 방법은 가장 위의 설명문을 참고하기 바랍니다.", "_____no_output_____" ] ] ]

[ "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "code" ], [ "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# 303 Build NN Quickly\n\nView more, visit my tutorial page: https://morvanzhou.github.io/tutorials/\nMy Youtube Channel: https://www.youtube.com/user/MorvanZhou", "_____no_output_____" ] ], [ [ "import torch\nimport torch.nn.functional as F", "_____no_output_____" ], [ "# replace following class code with an easy sequential network\nclass Net(torch.nn.Module):\n def __init__(self, n_feature, n_hidden, n_output):\n super(Net, self).__init__()\n self.hidden = torch.nn.Linear(n_feature, n_hidden) # hidden layer\n self.predict = torch.nn.Linear(n_hidden, n_output) # output layer\n\n def forward(self, x):\n x = F.relu(self.hidden(x)) # activation function for hidden layer\n x = self.predict(x) # linear output\n return x", "_____no_output_____" ], [ "net1 = Net(1, 10, 1)", "_____no_output_____" ], [ "# easy and fast way to build your network\nnet2 = torch.nn.Sequential(\n torch.nn.Linear(1, 10),\n torch.nn.ReLU(),\n torch.nn.Linear(10, 1)\n)\n", "_____no_output_____" ], [ "print(net1) # net1 architecture\nprint(net2) # net2 architecture", "Net(\n (hidden): Linear(in_features=1, out_features=10, bias=True)\n (predict): Linear(in_features=10, out_features=1, bias=True)\n)\nSequential(\n (0): Linear(in_features=1, out_features=10, bias=True)\n (1): ReLU()\n (2): Linear(in_features=10, out_features=1, bias=True)\n)\n" ] ] ]

[ "markdown", "code" ]

[ [ "markdown" ], [ "code", "code", "code", "code", "code" ] ]

[ "BSD-2-Clause" ]

[ "BSD-2-Clause" ]

[ "BSD-2-Clause" ]

[ [ [ "<script async src=\"https://www.googletagmanager.com/gtag/js?id=UA-59152712-8\"></script>\n<script>\n window.dataLayer = window.dataLayer || [];\n function gtag(){dataLayer.push(arguments);}\n gtag('js', new Date());\n\n gtag('config', 'UA-59152712-8');\n</script>\n\n# Generating C code for the right-hand-side of the scalar wave equation, in ***curvilinear*** coordinates, using a reference metric formalism\n\n## Author: Zach Etienne\n### Formatting improvements courtesy Brandon Clark\n\n[comment]: <> (Abstract: TODO)\n\n**Notebook Status:** <font color='green'><b> Validated </b></font>\n\n**Validation Notes:** This tutorial notebook has been confirmed to be self-consistent with its corresponding NRPy+ module, as documented [below](#code_validation). In addition, all expressions have been validated against a trusted code (the [original SENR/NRPy+ code](https://bitbucket.org/zach_etienne/nrpy)).\n\n### NRPy+ Source Code for this module: [ScalarWaveCurvilinear/ScalarWaveCurvilinear_RHSs.py](../edit/ScalarWaveCurvilinear/ScalarWaveCurvilinear_RHSs.py)\n\n[comment]: <> (Introduction: TODO)", "_____no_output_____" ], [ "<a id='toc'></a>\n\n# Table of Contents\n$$\\label{toc}$$\n\nThis notebook is organized as follows\n\n0. [Preliminaries](#prelim): Reference Metrics and Picking Best Coordinate System to Solve the PDE\n1. [Example](#example): The scalar wave equation in spherical coordinates\n1. [Step 1](#contracted_christoffel): Contracted Christoffel symbols $\\hat{\\Gamma}^i = \\hat{g}^{ij}\\hat{\\Gamma}^k_{ij}$ in spherical coordinates, using NRPy+\n1. [Step 2](#rhs_scalarwave_spherical): The right-hand side of the scalar wave equation in spherical coordinates, using NRPy+\n1. [Step 3](#code_validation): Code Validation against `ScalarWaveCurvilinear.ScalarWaveCurvilinear_RHSs` NRPy+ Module\n1. [Step 5](#latex_pdf_output): Output this notebook to $\\LaTeX$-formatted PDF file", "_____no_output_____" ], [ "<a id='prelim'></a>\n\n# Preliminaries: Reference Metrics and Picking Best Coordinate System to Solve the PDE \\[Back to [top](#toc)\\]\n$$\\label{prelim}$$\n\nRecall from [NRPy+ tutorial notebook on the Cartesian scalar wave equation](Tutorial-ScalarWave.ipynb), the scalar wave equation in 3D Cartesian coordinates is given by\n\n$$\\partial_t^2 u = c^2 \\nabla^2 u \\text{,}$$\nwhere $u$ (the amplitude of the wave) is a function of time and Cartesian coordinates in space: $u = u(t,x,y,z)$ (spatial dimension as-yet unspecified), and subject to some initial condition\n$$u(0,x,y,z) = f(x,y,z),$$\n\nwith suitable (sometimes approximate) spatial boundary conditions.\n\nTo simplify this equation, let's first choose units such that $c=1$. Alternative wave speeds can be constructed\nby simply rescaling the time coordinate, with the net effect being that the time $t$ is replaced with time in dimensions of space; i.e., $t\\to c t$:\n\n$$\\partial_t^2 u = \\nabla^2 u.$$\n\nAs we learned in the [NRPy+ tutorial notebook on reference metrics](Tutorial-Reference_Metric.ipynb), reference metrics are a means to pick the best coordinate system for the PDE we wish to solve. However, to take advantage of reference metrics requires first that we generalize the PDE. In the case of the scalar wave equation, this involves first rewriting in [Einstein notation](https://en.wikipedia.org/wiki/Einstein_notation) (with implied summation over repeated indices) via\n\n$$(-\\partial_t^2 + \\nabla^2) u = \\eta^{\\mu\\nu} u_{,\\ \\mu\\nu} = 0,$$\n\nwhere $u_{,\\mu\\nu} = \\partial_\\mu \\partial_\\nu u$, and $\\eta^{\\mu\\nu}$ is the contravariant flat-space metric tensor with components $\\text{diag}(-1,1,1,1)$.\n\nNext we apply the \"comma-goes-to-semicolon rule\" and replace $\\eta^{\\mu\\nu}$ with $\\hat{g}^{\\mu\\nu}$ to generalize the scalar wave equation to an arbitrary reference metric $\\hat{g}^{\\mu\\nu}$:\n\n$$\\hat{g}^{\\mu\\nu} u_{;\\ \\mu\\nu} = \\hat{\\nabla}_{\\mu} \\hat{\\nabla}_{\\nu} u = 0,$$\n\nwhere $\\hat{\\nabla}_{\\mu}$ denotes the [covariant derivative](https://en.wikipedia.org/wiki/Covariant_derivative) with respect to the reference metric basis vectors $\\hat{x}^{\\mu}$, and $\\hat{g}^{\\mu \\nu} \\hat{\\nabla}_{\\mu} \\hat{\\nabla}_{\\nu} u$ is the covariant\n[D'Alembertian](https://en.wikipedia.org/wiki/D%27Alembert_operator) of $u$.\n\nFor example, suppose we wish to model a short-wavelength wave that is nearly spherical. In this case, if we were to solve the wave equation PDE in Cartesian coordinates, we would in principle need high resolution in all three cardinal directions. If instead we chose spherical coordinates centered at the center of the wave, we might need high resolution only in the radial direction, with only a few points required in the angular directions. Thus choosing spherical coordinates would be far more computationally efficient than modeling the wave in Cartesian coordinates.\n\nLet's now expand the covariant scalar wave equation in arbitrary coordinates. Since the covariant derivative of a scalar is equivalent to its partial derivative, we have\n\\begin{align}\n0 &= \\hat{g}^{\\mu \\nu} \\hat{\\nabla}_{\\mu} \\hat{\\nabla}_{\\nu} u \\\\\n&= \\hat{g}^{\\mu \\nu} \\hat{\\nabla}_{\\mu} \\partial_{\\nu} u.\n\\end{align}\n\n$\\partial_{\\nu} u$ transforms as a one-form under covariant differentiation, so we have\n$$\\hat{\\nabla}_{\\mu} \\partial_{\\nu} u = \\partial_{\\mu} \\partial_{\\nu} u - \\hat{\\Gamma}^\\tau_{\\mu\\nu} \\partial_\\tau u,$$\nwhere \n\n$$\\hat{\\Gamma}^\\tau_{\\mu\\nu} = \\frac{1}{2} \\hat{g}^{\\tau\\alpha} \\left(\\partial_\\nu \\hat{g}_{\\alpha\\mu} + \\partial_\\mu \\hat{g}_{\\alpha\\nu} - \\partial_\\alpha \\hat{g}_{\\mu\\nu} \\right)$$\nare the [Christoffel symbols](https://en.wikipedia.org/wiki/Christoffel_symbols) associated with the reference metric $\\hat{g}_{\\mu\\nu}$.\n\nThen the scalar wave equation is written:\n$$0 = \\hat{g}^{\\mu \\nu} \\left( \\partial_{\\mu} \\partial_{\\nu} u - \\hat{\\Gamma}^\\tau_{\\mu\\nu} \\partial_\\tau u\\right).$$\n\nDefine the contracted Christoffel symbols:\n$$\\hat{\\Gamma}^\\tau = \\hat{g}^{\\mu\\nu} \\hat{\\Gamma}^\\tau_{\\mu\\nu}.$$\n\nThen the scalar wave equation is given by\n$$0 = \\hat{g}^{\\mu \\nu} \\partial_{\\mu} \\partial_{\\nu} u - \\hat{\\Gamma}^\\tau \\partial_\\tau u.$$\n\nThe reference metrics we adopt satisfy\n$$\\hat{g}^{t \\nu} = -\\delta^{t \\nu},$$\nwhere $\\delta^{t \\nu}$ is the [Kronecker delta](https://en.wikipedia.org/wiki/Kronecker_delta). Therefore the scalar wave equation in curvilinear coordinates can be written\n\\begin{align}\n0 &= \\hat{g}^{\\mu \\nu} \\partial_{\\mu} \\partial_{\\nu} u - \\hat{\\Gamma}^\\tau \\partial_\\tau u \\\\\n&= -\\partial_t^2 u + \\hat{g}^{i j} \\partial_{i} \\partial_{j} u - \\hat{\\Gamma}^i \\partial_i u \\\\\n\\implies \\partial_t^2 u &= \\hat{g}^{i j} \\partial_{i} \\partial_{j} u - \\hat{\\Gamma}^i \\partial_i u,\n\\end{align}\nwhere repeated Latin indices denote implied summation over *spatial* components only. This module implements the bottom equation for arbitrary reference metrics satisfying $\\hat{g}^{t \\nu} = -\\delta^{t \\nu}$. To gain an appreciation for what NRPy+ accomplishes automatically, let's first work out the scalar wave equation in spherical coordinates by hand:", "_____no_output_____" ], [ "<a id='example'></a>\n\n# Example: The scalar wave equation in spherical coordinates \\[Back to [top](#toc)\\]\n$$\\label{example}$$\n\nFor example, the spherical reference metric is written\n\n$$\\hat{g}_{\\mu\\nu} = \\begin{pmatrix}\n-1 & 0 & 0 & 0 \\\\\n 0 & 1 & 0 & 0 \\\\\n 0 & 0 & r^2 & 0 \\\\\n 0 & 0 & 0 & r^2 \\sin^2 \\theta \\\\\n\\end{pmatrix}.\n$$\n\nSince the inverse of a diagonal matrix is simply the inverse of the diagonal elements, we can write \n$$\\hat{g}^{\\mu\\nu} = \\begin{pmatrix}\n-1 & 0 & 0 & 0 \\\\\n 0 & 1 & 0 & 0 \\\\\n 0 & 0 & \\frac{1}{r^2} & 0 \\\\\n 0 & 0 & 0 & \\frac{1}{r^2 \\sin^2 \\theta} \\\\\n\\end{pmatrix}.$$\n\nThe scalar wave equation in these coordinates can thus be written\n\\begin{align}\n0 &= \\hat{g}^{\\mu \\nu} \\partial_{\\mu} \\partial_{\\nu} u - \\hat{\\Gamma}^\\tau \\partial_\\tau u \\\\\n&= \\hat{g}^{tt} \\partial_t^2 u + \\hat{g}^{rr} \\partial_r^2 u + \\hat{g}^{\\theta\\theta} \\partial_\\theta^2 u + \\hat{g}^{\\phi\\phi} \\partial_\\phi^2 u - \\hat{\\Gamma}^\\tau \\partial_\\tau u \\\\\n&= -\\partial_t^2 u + \\partial_r^2 u + \\frac{1}{r^2} \\partial_\\theta^2\nu + \\frac{1}{r^2 \\sin^2 \\theta} \\partial_\\phi^2 u - \\hat{\\Gamma}^\\tau \\partial_\\tau u\\\\\n\\implies \\partial_t^2 u &= \\partial_r^2 u + \\frac{1}{r^2} \\partial_\\theta^2\nu + \\frac{1}{r^2 \\sin^2 \\theta} \\partial_\\phi^2 u - \\hat{\\Gamma}^\\tau \\partial_\\tau u\n\\end{align}\n\nThe contracted Christoffel symbols \n$\\hat{\\Gamma}^\\tau$ can then be computed directly from the metric $\\hat{g}_{\\mu\\nu}$.\n\nIt can be shown (exercise to the reader) that the only nonzero\ncomponents of $\\hat{\\Gamma}^\\tau$ in static spherical polar coordinates are\ngiven by\n\\begin{align}\n\\hat{\\Gamma}^r &= -\\frac{2}{r} \\\\\n\\hat{\\Gamma}^\\theta &= -\\frac{\\cos\\theta}{r^2 \\sin\\theta}.\n\\end{align}\n\nThus we have found the Laplacian in spherical coordinates is simply:\n\n\\begin{align}\n\\nabla^2 u &= \n\\partial_r^2 u + \\frac{1}{r^2} \\partial_\\theta^2 u + \\frac{1}{r^2 \\sin^2 \\theta} \\partial_\\phi^2 u - \\hat{\\Gamma}^\\tau \\partial_\\tau u\\\\\n&= \\partial_r^2 u + \\frac{1}{r^2} \\partial_\\theta^2 u + \\frac{1}{r^2 \\sin^2 \\theta} \\partial_\\phi^2 u + \\frac{2}{r} \\partial_r u + \\frac{\\cos\\theta}{r^2 \\sin\\theta} \\partial_\\theta u\n\\end{align}\n(cf. http://mathworld.wolfram.com/SphericalCoordinates.html; though note that they defined the angle $\\phi$ as $\\theta$ and $\\theta$ as $\\phi$.)", "_____no_output_____" ], [ "<a id='contracted_christoffel'></a>\n\n# Step 1: Contracted Christoffel symbols $\\hat{\\Gamma}^i = \\hat{g}^{ij}\\hat{\\Gamma}^k_{ij}$ in spherical coordinates, using NRPy+ \\[Back to [top](#toc)\\]\n$$\\label{contracted_christoffel}$$\n\nLet's next use NRPy+ to derive the contracted Christoffel symbols\n$$\\hat{g}^{ij} \\hat{\\Gamma}^k_{ij}$$\nin spherical coordinates, where $i\\in\\{1,2,3\\}$ and $j\\in\\{1,2,3\\}$ are spatial indices.\n\nAs discussed in the [NRPy+ tutorial notebook on reference metrics](Tutorial-Reference_Metric.ipynb), several reference-metric-related quantities in spherical coordinates are computed in NRPy+ (provided the parameter **`reference_metric::CoordSystem`** is set to **`\"Spherical\"`**), including the inverse spatial spherical reference metric $\\hat{g}^{ij}$ and the Christoffel symbols from this reference metric $\\hat{\\Gamma}^{i}_{jk}$. ", "_____no_output_____" ] ], [ [ "import sympy as sp\nimport NRPy_param_funcs as par\nimport indexedexp as ixp\nimport reference_metric as rfm\n\n# reference_metric::CoordSystem can be set to Spherical, SinhSpherical, SinhSphericalv2, \n# Cylindrical, SinhCylindrical, SinhCylindricalv2, etc.\n# See reference_metric.py and NRPy+ tutorial notebook on \n# reference metrics for full list and description of how\n# to extend.\npar.set_parval_from_str(\"reference_metric::CoordSystem\",\"Spherical\")\npar.set_parval_from_str(\"grid::DIM\",3)\n\nrfm.reference_metric()\n\ncontractedGammahatU = ixp.zerorank1()\nfor k in range(3):\n for i in range(3):\n for j in range(3):\n contractedGammahatU[k] += rfm.ghatUU[i][j] * rfm.GammahatUDD[k][i][j]\n\nfor k in range(3):\n print(\"contracted GammahatU[\"+str(k)+\"]:\")\n sp.pretty_print(sp.simplify(contractedGammahatU[k]))\n if k<2:\n print(\"\\n\\n\")", "contracted GammahatU[0]:\n-2 \n───\nxx₀\n\n\n\ncontracted GammahatU[1]:\n -1 \n─────────────\n 2 \nxx₀ ⋅tan(xx₁)\n\n\n\ncontracted GammahatU[2]:\n0\n" ] ], [ [ "<a id='rhs_scalarwave_spherical'></a>\n\n# Step 2: The right-hand side of the scalar wave equation in spherical coordinates, using NRPy+ \\[Back to [top](#toc)\\]\n$$\\label{rhs_scalarwave_spherical}$$\n\nFollowing our [implementation of the scalar wave equation in Cartesian coordinates](Tutorial-ScalarWave.ipynb), we will introduce a new variable $v=\\partial_t u$ that will enable us to split the second time derivative into two first-order time derivatives:\n\n\\begin{align}\n\\partial_t u &= v \\\\\n\\partial_t v &= \\hat{g}^{ij} \\partial_{i} \\partial_{j} u - \\hat{\\Gamma}^i \\partial_i u.\n\\end{align}\n\nAdding back the sound speed $c$, we have a choice of a single factor of $c$ multiplying both right-hand sides, or a factor of $c^2$ multiplying the second equation only. We'll choose the latter:\n\n\\begin{align}\n\\partial_t u &= v \\\\\n\\partial_t v &= c^2 \\left(\\hat{g}^{ij} \\partial_{i} \\partial_{j} u - \\hat{\\Gamma}^i \\partial_i u\\right).\n\\end{align}\n\nNow let's generate the C code for the finite-difference representations of the right-hand sides of the above \"time evolution\" equations for $u$ and $v$. Since the right-hand side of $\\partial_t v$ contains implied sums over $i$ and $j$ in the first term, and an implied sum over $k$ in the second term, we'll find it useful to split the right-hand side into two parts\n\n\\begin{equation}\n\\partial_t v = c^2 \\left(\n{\\underbrace {\\textstyle \\hat{g}^{ij} \\partial_{i} \\partial_{j} u}_{\\text{Part 1}}} \n{\\underbrace {\\textstyle -\\hat{\\Gamma}^i \\partial_i u}_{\\text{Part 2}}}\\right),\n\\end{equation}\n\nand perform the implied sums in two pieces:", "_____no_output_____" ] ], [ [ "import NRPy_param_funcs as par\nimport indexedexp as ixp\nimport grid as gri\nimport finite_difference as fin\nimport reference_metric as rfm\nfrom outputC import *", "_____no_output_____" ], [ "# The name of this module (\"scalarwave\") is given by __name__:\nthismodule = __name__\n\n# Step 0: Read the spatial dimension parameter as DIM.\nDIM = par.parval_from_str(\"grid::DIM\")\n\n# Step 1: Set the finite differencing order to 4.\npar.set_parval_from_str(\"finite_difference::FD_CENTDERIVS_ORDER\",4)\n\n# Step 2a: Reset the gridfunctions list; below we define the\n# full complement of gridfunctions needed by this\n# tutorial. This line of code enables us to re-run this\n# tutorial without resetting the running Python kernel.\ngri.glb_gridfcs_list = []\n# Step 2b: Register gridfunctions that are needed as input\n# to the scalar wave RHS expressions.\nuu, vv = gri.register_gridfunctions(\"EVOL\",[\"uu\",\"vv\"])\n\n# Step 3a: Declare the rank-1 indexed expression \\partial_{i} u,\n# Derivative variables like these must have an underscore\n# in them, so the finite difference module can parse the\n# variable name properly.\nuu_dD = ixp.declarerank1(\"uu_dD\")\n\n# Step 3b: Declare the rank-2 indexed expression \\partial_{ij} u,\n# which is symmetric about interchange of indices i and j\n# Derivative variables like these must have an underscore\n# in them, so the finite difference module can parse the\n# variable name properly.\nuu_dDD = ixp.declarerank2(\"uu_dDD\",\"sym01\")\n\n# Step 4: Define the C parameter wavespeed. The `wavespeed`\n# variable is a proper SymPy variable, so it can be\n# used in below expressions. In the C code, it acts\n# just like a usual parameter, whose value is \n# specified in the parameter file.\nwavespeed = par.Cparameters(\"REAL\",thismodule,\"wavespeed\", 1.0)\n\n# Step 5: Define right-hand sides for the evolution.\nuu_rhs = vv\n# Step 5b: The right-hand side of the \\partial_t v equation\n# is given by:\n# \\hat{g}^{ij} \\partial_i \\partial_j u - \\hat{\\Gamma}^i \\partial_i u.\n# ^^^^^^^^^^^^ PART 1 ^^^^^^^^^^^^^^^^ ^^^^^^^^^^ PART 2 ^^^^^^^^^^^\nvv_rhs = 0\nfor i in range(DIM):\n # PART 2:\n vv_rhs -= contractedGammahatU[i]*uu_dD[i]\n for j in range(DIM):\n # PART 1:\n vv_rhs += rfm.ghatUU[i][j]*uu_dDD[i][j]\n\nvv_rhs *= wavespeed*wavespeed\n\n# Step 6: Generate C code for scalarwave evolution equations,\n# print output to the screen (standard out, or stdout).\nfin.FD_outputC(\"stdout\",\n [lhrh(lhs=gri.gfaccess(\"rhs_gfs\",\"uu\"),rhs=uu_rhs),\n lhrh(lhs=gri.gfaccess(\"rhs_gfs\",\"vv\"),rhs=vv_rhs)])", "{\n /* \n * NRPy+ Finite Difference Code Generation, Step 1 of 2: Read from main memory and compute finite difference stencils:\n */\n /*\n * Original SymPy expressions:\n * \"[const double uu_dD0 = invdx0*(-2*uu_i0m1_i1_i2/3 + uu_i0m2_i1_i2/12 + 2*uu_i0p1_i1_i2/3 - uu_i0p2_i1_i2/12),\n * const double uu_dD1 = invdx1*(-2*uu_i0_i1m1_i2/3 + uu_i0_i1m2_i2/12 + 2*uu_i0_i1p1_i2/3 - uu_i0_i1p2_i2/12),\n * const double uu_dDD00 = invdx0**2*(-5*uu/2 + 4*uu_i0m1_i1_i2/3 - uu_i0m2_i1_i2/12 + 4*uu_i0p1_i1_i2/3 - uu_i0p2_i1_i2/12),\n * const double uu_dDD11 = invdx1**2*(-5*uu/2 + 4*uu_i0_i1m1_i2/3 - uu_i0_i1m2_i2/12 + 4*uu_i0_i1p1_i2/3 - uu_i0_i1p2_i2/12),\n * const double uu_dDD22 = invdx2**2*(-5*uu/2 + 4*uu_i0_i1_i2m1/3 - uu_i0_i1_i2m2/12 + 4*uu_i0_i1_i2p1/3 - uu_i0_i1_i2p2/12)]\"\n */\n const double uu_i0_i1_i2m2 = in_gfs[IDX4(UUGF, i0,i1,i2-2)];\n const double uu_i0_i1_i2m1 = in_gfs[IDX4(UUGF, i0,i1,i2-1)];\n const double uu_i0_i1m2_i2 = in_gfs[IDX4(UUGF, i0,i1-2,i2)];\n const double uu_i0_i1m1_i2 = in_gfs[IDX4(UUGF, i0,i1-1,i2)];\n const double uu_i0m2_i1_i2 = in_gfs[IDX4(UUGF, i0-2,i1,i2)];\n const double uu_i0m1_i1_i2 = in_gfs[IDX4(UUGF, i0-1,i1,i2)];\n const double uu = in_gfs[IDX4(UUGF, i0,i1,i2)];\n const double uu_i0p1_i1_i2 = in_gfs[IDX4(UUGF, i0+1,i1,i2)];\n const double uu_i0p2_i1_i2 = in_gfs[IDX4(UUGF, i0+2,i1,i2)];\n const double uu_i0_i1p1_i2 = in_gfs[IDX4(UUGF, i0,i1+1,i2)];\n const double uu_i0_i1p2_i2 = in_gfs[IDX4(UUGF, i0,i1+2,i2)];\n const double uu_i0_i1_i2p1 = in_gfs[IDX4(UUGF, i0,i1,i2+1)];\n const double uu_i0_i1_i2p2 = in_gfs[IDX4(UUGF, i0,i1,i2+2)];\n const double vv = in_gfs[IDX4(VVGF, i0,i1,i2)];\n const double tmpFD0 = (1.0/12.0)*uu_i0m2_i1_i2;\n const double tmpFD1 = -1.0/12.0*uu_i0p2_i1_i2;\n const double tmpFD2 = (1.0/12.0)*uu_i0_i1m2_i2;\n const double tmpFD3 = -1.0/12.0*uu_i0_i1p2_i2;\n const double tmpFD4 = -5.0/2.0*uu;\n const double uu_dD0 = invdx0*(tmpFD0 + tmpFD1 - 2.0/3.0*uu_i0m1_i1_i2 + (2.0/3.0)*uu_i0p1_i1_i2);\n const double uu_dD1 = invdx1*(tmpFD2 + tmpFD3 - 2.0/3.0*uu_i0_i1m1_i2 + (2.0/3.0)*uu_i0_i1p1_i2);\n const double uu_dDD00 = ((invdx0)*(invdx0))*(-tmpFD0 + tmpFD1 + tmpFD4 + (4.0/3.0)*uu_i0m1_i1_i2 + (4.0/3.0)*uu_i0p1_i1_i2);\n const double uu_dDD11 = ((invdx1)*(invdx1))*(-tmpFD2 + tmpFD3 + tmpFD4 + (4.0/3.0)*uu_i0_i1m1_i2 + (4.0/3.0)*uu_i0_i1p1_i2);\n const double uu_dDD22 = ((invdx2)*(invdx2))*(tmpFD4 + (4.0/3.0)*uu_i0_i1_i2m1 - 1.0/12.0*uu_i0_i1_i2m2 + (4.0/3.0)*uu_i0_i1_i2p1 - 1.0/12.0*uu_i0_i1_i2p2);\n /* \n * NRPy+ Finite Difference Code Generation, Step 2 of 2: Evaluate SymPy expressions and write to main memory:\n */\n /*\n * Original SymPy expressions:\n * \"[rhs_gfs[IDX4(UUGF, i0, i1, i2)] = vv,\n * rhs_gfs[IDX4(VVGF, i0, i1, i2)] = wavespeed**2*(2*uu_dD0/xx0 + uu_dD1*sin(2*xx1)/(2*xx0**2*sin(xx1)**2) + uu_dDD00 + uu_dDD11/xx0**2 + uu_dDD22/(xx0**2*sin(xx1)**2))]\"\n */\n const double tmp0 = (1.0/((xx0)*(xx0)));\n const double tmp1 = tmp0/((sin(xx1))*(sin(xx1)));\n rhs_gfs[IDX4(UUGF, i0, i1, i2)] = vv;\n rhs_gfs[IDX4(VVGF, i0, i1, i2)] = ((wavespeed)*(wavespeed))*(tmp0*uu_dDD11 + (1.0/2.0)*tmp1*uu_dD1*sin(2*xx1) + tmp1*uu_dDD22 + 2*uu_dD0/xx0 + uu_dDD00);\n}\n\n" ] ], [ [ "<a id='code_validation'></a>\n\n# Step 3: Code Validation against `ScalarWaveCurvilinear.ScalarWaveCurvilinear_RHSs` NRPy+ Module \\[Back to [top](#toc)\\]\n$$\\label{code_validation}$$\n\nHere, as a code validation check, we verify agreement in the SymPy expressions for the RHSs of the Curvilinear Scalar Wave equation (i.e., uu_rhs and vv_rhs) between\n\n1. this tutorial and \n2. the NRPy+ [ScalarWaveCurvilinear.ScalarWaveCurvilinear_RHSs](../edit/ScalarWaveCurvilinear/ScalarWaveCurvilinear_RHSs.py) module.\n\nBy default, we analyze the RHSs in Spherical coordinates, though other coordinate systems may be chosen.", "_____no_output_____" ] ], [ [ "# Step 7: We already have SymPy expressions for uu_rhs and vv_rhs in\n# terms of other SymPy variables. Even if we reset the list\n# of NRPy+ gridfunctions, these *SymPy* expressions for\n# uu_rhs and vv_rhs *will remain unaffected*. \n# \n# Here, we will use the above-defined uu_rhs and vv_rhs to \n# validate against the same expressions in the \n# ScalarWaveCurvilinear/ScalarWaveCurvilinear module,\n# to ensure consistency between the tutorial and the\n# module itself.\n#\n# Reset the list of gridfunctions, as registering a gridfunction\n# twice will spawn an error.\ngri.glb_gridfcs_list = []\n\n# Step 8: Call the ScalarWaveCurvilinear_RHSs() function from within the\n# ScalarWaveCurvilinear/ScalarWaveCurvilinear_RHSs.py module,\n# which should do exactly the same as in Steps 1-6 above.\nimport ScalarWaveCurvilinear.ScalarWaveCurvilinear_RHSs as swcrhs\nswcrhs.ScalarWaveCurvilinear_RHSs()\n\n# Step 9: Consistency check between the tutorial notebook above\n# and the ScalarWaveCurvilinear_RHSs() function from within the\n# ScalarWaveCurvilinear/ScalarWaveCurvilinear_RHSs.py module.\nprint(\"Consistency check between ScalarWaveCurvilinear tutorial and NRPy+ module:\")\nprint(\"uu_rhs - swcrhs.uu_rhs: \"+str(sp.simplify(uu_rhs - swcrhs.uu_rhs))+\"\\t\\t (should be zero)\")\nprint(\"vv_rhs - swcrhs.vv_rhs: \"+str(sp.simplify(vv_rhs - swcrhs.vv_rhs))+\"\\t\\t (should be zero)\")", "Consistency check between ScalarWaveCurvilinear tutorial and NRPy+ module:\nuu_rhs - swcrhs.uu_rhs: 0\t\t (should be zero)\nvv_rhs - swcrhs.vv_rhs: 0\t\t (should be zero)\n" ] ], [ [ "<a id='latex_pdf_output'></a>\n\n# Step 4: Output this notebook to $\\LaTeX$-formatted PDF file \\[Back to [top](#toc)\\]\n$$\\label{latex_pdf_output}$$\n\nThe following code cell converts this Jupyter notebook into a proper, clickable $\\LaTeX$-formatted PDF file. After the cell is successfully run, the generated PDF may be found in the root NRPy+ tutorial directory, with filename\n[Tutorial-ScalarWaveCurvilinear.pdf](Tutorial-ScalarWaveCurvilinear.pdf) (Note that clicking on this link may not work; you may need to open the PDF file through another means.)", "_____no_output_____" ] ], [ [ "!jupyter nbconvert --to latex --template latex_nrpy_style.tplx --log-level='WARN' Tutorial-ScalarWaveCurvilinear.ipynb\n!pdflatex -interaction=batchmode Tutorial-ScalarWaveCurvilinear.tex\n!pdflatex -interaction=batchmode Tutorial-ScalarWaveCurvilinear.tex\n!pdflatex -interaction=batchmode Tutorial-ScalarWaveCurvilinear.tex\n!rm -f Tut*.out Tut*.aux Tut*.log", "[pandoc warning] Duplicate link reference `[comment]' \"source\" (line 23, column 1)\nThis is pdfTeX, Version 3.14159265-2.6-1.40.18 (TeX Live 2017/Debian) (preloaded format=pdflatex)\n restricted \\write18 enabled.\nentering extended mode\nThis is pdfTeX, Version 3.14159265-2.6-1.40.18 (TeX Live 2017/Debian) (preloaded format=pdflatex)\n restricted \\write18 enabled.\nentering extended mode\nThis is pdfTeX, Version 3.14159265-2.6-1.40.18 (TeX Live 2017/Debian) (preloaded format=pdflatex)\n restricted \\write18 enabled.\nentering extended mode\n" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ [ [ "import os\nimport urllib\nfrom zipfile import ZipFile\nimport fileinput\nimport numpy as np\nimport gc\nimport urllib.request", "_____no_output_____" ], [ "if not os.path.exists('glove.840B.300d.txt'):\n if not os.path.exists('glove.840B.300d.zip'):\n print('downloading GloVe')\n urllib.request.urlretrieve(\"http://nlp.stanford.edu/data/glove.840B.300d.zip\", \"glove.840B.300d.zip\")\n zip = ZipFile('glove.840B.300d.zip')\n zip.extractall()", "_____no_output_____" ], [ "import torch\nfrom torchtext import data\nfrom torchtext import datasets\nfrom torchtext.vocab import GloVe\nimport fileinput\nimport numpy as np\nfrom cove import MTLSTM\n\ninputs = data.Field(lower=True, include_lengths=True, batch_first=True)\nanswers = data.Field(sequential=False)\n\nprint('Generating train, dev, test splits')\ntrain, dev, test = datasets.SNLI.splits(inputs, answers)\n\nprint('Building vocabulary')\ninputs.build_vocab(train, dev, test)\n\ng = GloVe(name='840B', dim=300)\ngc.collect()\ninputs.vocab.load_vectors(vectors=g)\ngc.collect()\n\nanswers.build_vocab(train)\n\nmodel = MTLSTM(n_vocab=len(inputs.vocab), vectors=inputs.vocab.vectors)\nmodel.cuda(0)\n\ntrain_iter, dev_iter, test_iter = data.BucketIterator.splits(\n (train, dev, test), batch_size=100, device=0)\n\ntrain_iter.init_epoch()", "Generating train, dev, test splits\nBuilding vocabulary\n" ], [ "from keras.models import load_model\nimport tensorflow as tf", "/usr/local/lib/python3.5/dist-packages/h5py/__init__.py:34: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\n from ._conv import register_converters as _register_converters\nUsing TensorFlow backend.\n" ], [ "# To prevent Tensorflow from being greedy and allocating all GPU memory for itself\ngpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.333)\nsess = tf.Session(config=tf.ConfigProto(gpu_options=gpu_options))\n\n# Loading loding saved Keras CoVe model\ncove_model = load_model('Keras_CoVe.h5')", "/usr/local/lib/python3.5/dist-packages/keras/models.py:255: UserWarning: No training configuration found in save file: the model was *not* compiled. Compile it manually.\n warnings.warn('No training configuration found in save file: '\n" ], [ "TOTAL_NUM_TEST_SENTENCE = 10000\nprint('Comparing Keras CoVe prediction with Pytorch CoVe')\nabs_error_per_dim = 0\ntotal_num_of_dim = 0\nnum_test_sentence = 0\nmodel.train()\nfor batch_idx, batch in enumerate(train_iter):\n if num_test_sentence > TOTAL_NUM_TEST_SENTENCE:\n # It takes a long time to run through all examples hence restricting the test set \n break\n cove_premise = model(*batch.premise)\n #cove_hypothesis = model(*batch.hypothesis)\n sentence_sparse_vector = batch.premise[0].data.cpu().numpy()\n for i in range(len(sentence_sparse_vector)):\n sentence = sentence_sparse_vector[i]\n sentence_glove = []\n for word in sentence:\n sentence_glove.append(inputs.vocab.vectors[word].numpy())\n sentence_glove = np.expand_dims(np.array(sentence_glove),0)\n if np.any(np.sum(sentence_glove,axis=2)==0):\n break\n keras_cove_sentence = cove_model.predict(sentence_glove)\n keras_cove_sentence = np.squeeze(keras_cove_sentence,0)\n pytorch_cove_sentence = cove_premise.data.cpu().numpy()[i]\n\n abs_error_per_dim+=np.sum(np.abs(keras_cove_sentence - pytorch_cove_sentence))\n total_num_of_dim+=np.prod(sentence_glove.shape)\n num_test_sentence+=1\nabs_error_per_dim/=total_num_of_dim\nprint('abs error per dim:'+str(abs_error_per_dim))", "Comparing Keras CoVe prediction with Pytorch CoVe\nabs error per dim:2.4579888586353267e-08\n" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]