Datasets:

---

bigcode

/

jupyter-parsed

like 4

Follow

BigCode 1.34k

[ "CC0-1.0" ]

[ "CC0-1.0" ]

[ "CC0-1.0" ]

[ [ [ "#@title Copyright 2019 Google LLC. { display-mode: \"form\" }\n# 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_____" ] ], [ [ "<table class=\"ee-notebook-buttons\" align=\"left\"><td>\n<a target=\"_blank\" href=\"http://colab.research.google.com/github/google/earthengine-api/blob/master/python/examples/ipynb/ee-api-colab-setup.ipynb\">\n <img src=\"https://www.tensorflow.org/images/colab_logo_32px.png\" /> Run in Google Colab</a>\n</td><td>\n<a target=\"_blank\" href=\"https://github.com/google/earthengine-api/blob/master/python/examples/ipynb/ee-api-colab-setup.ipynb\"><img width=32px src=\"https://www.tensorflow.org/images/GitHub-Mark-32px.png\" /> View source on GitHub</a></td></table>", "_____no_output_____" ], [ "# Earth Engine Python API Colab Setup\n\nThis notebook demonstrates how to setup the Earth Engine Python API in Colab and provides several examples of how to print and visualize Earth Engine processed data.\n\n## Import API and get credentials\n\nThe Earth Engine API is installed by default in Google Colaboratory so requires only importing and authenticating. These steps must be completed for each new Colab session, if you restart your Colab kernel, or if your Colab virtual machine is recycled due to inactivity.\n\n### Import the API\n\nRun the following cell to import the API into your session.", "_____no_output_____" ] ], [ [ "import ee", "_____no_output_____" ] ], [ [ "### Authenticate and initialize\n\nRun the `ee.Authenticate` function to authenticate your access to Earth Engine servers and `ee.Initialize` to initialize it. Upon running the following cell you'll be asked to grant Earth Engine access to your Google account. Follow the instructions printed to the cell.", "_____no_output_____" ] ], [ [ "# Trigger the authentication flow.\nee.Authenticate()\n\n# Initialize the library.\nee.Initialize()", "_____no_output_____" ] ], [ [ "### Test the API\n\nTest the API by printing the elevation of Mount Everest.", "_____no_output_____" ] ], [ [ "# Print the elevation of Mount Everest.\ndem = ee.Image('USGS/SRTMGL1_003')\nxy = ee.Geometry.Point([86.9250, 27.9881])\nelev = dem.sample(xy, 30).first().get('elevation').getInfo()\nprint('Mount Everest elevation (m):', elev)", "_____no_output_____" ] ], [ [ "## Map visualization\n\n`ee.Image` objects can be displayed to notebook output cells. The following two\nexamples demonstrate displaying a static image and an interactive map.\n", "_____no_output_____" ], [ "### Static image\n\nThe `IPython.display` module contains the `Image` function, which can display\nthe results of a URL representing an image generated from a call to the Earth\nEngine `getThumbUrl` function. The following cell will display a thumbnail\nof the global elevation model.", "_____no_output_____" ] ], [ [ "# Import the Image function from the IPython.display module. \nfrom IPython.display import Image\n\n# Display a thumbnail of global elevation.\nImage(url = dem.updateMask(dem.gt(0))\n .getThumbURL({'min': 0, 'max': 4000, 'dimensions': 512,\n 'palette': ['006633', 'E5FFCC', '662A00', 'D8D8D8', 'F5F5F5']}))", "_____no_output_____" ] ], [ [ "### Interactive map\n\nThe [`folium`](https://python-visualization.github.io/folium/)\nlibrary can be used to display `ee.Image` objects on an interactive\n[Leaflet](https://leafletjs.com/) map. Folium has no default\nmethod for handling tiles from Earth Engine, so one must be defined\nand added to the `folium.Map` module before use.\n\nThe following cell provides an example of adding a method for handing Earth Engine\ntiles and using it to display an elevation model to a Leaflet map.", "_____no_output_____" ] ], [ [ "# Import the Folium library.\nimport folium\n\n# Define a method for displaying Earth Engine image tiles to folium map.\ndef add_ee_layer(self, ee_image_object, vis_params, name):\n map_id_dict = ee.Image(ee_image_object).getMapId(vis_params)\n folium.raster_layers.TileLayer(\n tiles = map_id_dict['tile_fetcher'].url_format,\n attr = 'Map Data &copy; <a href=\"https://earthengine.google.com/\">Google Earth Engine</a>',\n name = name,\n overlay = True,\n control = True\n ).add_to(self)\n\n# Add EE drawing method to folium.\nfolium.Map.add_ee_layer = add_ee_layer\n\n# Set visualization parameters.\nvis_params = {\n 'min': 0,\n 'max': 4000,\n 'palette': ['006633', 'E5FFCC', '662A00', 'D8D8D8', 'F5F5F5']}\n\n# Create a folium map object.\nmy_map = folium.Map(location=[20, 0], zoom_start=3, height=500)\n\n# Add the elevation model to the map object.\nmy_map.add_ee_layer(dem.updateMask(dem.gt(0)), vis_params, 'DEM')\n\n# Add a layer control panel to the map.\nmy_map.add_child(folium.LayerControl())\n\n# Display the map.\ndisplay(my_map)", "_____no_output_____" ] ], [ [ "## Chart visualization\n\nSome Earth Engine functions produce tabular data that can be plotted by\ndata visualization packages such as `matplotlib`. The following example\ndemonstrates the display of tabular data from Earth Engine as a scatter\nplot. See [Charting in Colaboratory](https://colab.sandbox.google.com/notebooks/charts.ipynb)\nfor more information.", "_____no_output_____" ] ], [ [ "# Import the matplotlib.pyplot module.\nimport matplotlib.pyplot as plt\n\n# Fetch a Landsat image.\nimg = ee.Image('LANDSAT/LT05/C01/T1_SR/LT05_034033_20000913')\n\n# Select Red and NIR bands, scale them, and sample 500 points.\nsamp_fc = img.select(['B3','B4']).divide(10000).sample(scale=30, numPixels=500)\n\n# Arrange the sample as a list of lists.\nsamp_dict = samp_fc.reduceColumns(ee.Reducer.toList().repeat(2), ['B3', 'B4'])\nsamp_list = ee.List(samp_dict.get('list'))\n\n# Save server-side ee.List as a client-side Python list.\nsamp_data = samp_list.getInfo()\n\n# Display a scatter plot of Red-NIR sample pairs using matplotlib.\nplt.scatter(samp_data[0], samp_data[1], alpha=0.2)\nplt.xlabel('Red', fontsize=12)\nplt.ylabel('NIR', fontsize=12)\nplt.show()", "_____no_output_____" ] ] ]

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

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

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import numpy as np\nimport pandas as pd\n\nimport torch\nimport torchvision\nfrom torch.utils.data import Dataset, DataLoader\nfrom torchvision import transforms, utils\nimport torch.nn as nn\nimport torch.nn.functional as F\nimport torch.optim as optim\n\nfrom matplotlib import pyplot as plt\nfrom myrmsprop import MyRmsprop\n%matplotlib inline\ntorch.backends.cudnn.deterministic = True\ntorch.backends.cudnn.benchmark = False", "_____no_output_____" ], [ "train_data = np.load(\"train_type4_data.npy\",allow_pickle=True)\n\ntest_data = np.load(\"test_type4_data.npy\",allow_pickle=True)", "_____no_output_____" ], [ "mosaic_list_of_images = train_data[0][\"mosaic_list\"]\nmosaic_label = train_data[0][\"mosaic_label\"]\nfore_idx = train_data[0][\"fore_idx\"]\n\n\ntest_mosaic_list_of_images = test_data[0][\"mosaic_list\"]\ntest_mosaic_label = test_data[0][\"mosaic_label\"]\ntest_fore_idx = test_data[0][\"fore_idx\"]", "_____no_output_____" ], [ "class MosaicDataset1(Dataset):\n \"\"\"MosaicDataset dataset.\"\"\"\n\n def __init__(self, mosaic_list, mosaic_label,fore_idx):\n \"\"\"\n Args:\n csv_file (string): Path to the csv file with annotations.\n root_dir (string): Directory with all the images.\n transform (callable, optional): Optional transform to be applied\n on a sample.\n \"\"\"\n self.mosaic = mosaic_list\n self.label = mosaic_label\n self.fore_idx = fore_idx\n \n def __len__(self):\n return len(self.label)\n\n def __getitem__(self, idx):\n return self.mosaic[idx] , self.label[idx] , self.fore_idx[idx]", "_____no_output_____" ], [ "batch = 3000\ntrain_dataset = MosaicDataset1(mosaic_list_of_images, mosaic_label, fore_idx)\ntrain_loader = DataLoader( train_dataset,batch_size= batch ,shuffle=False)\n#batch = 2000\n#test_dataset = MosaicDataset1(test_mosaic_list_of_images, test_mosaic_label, test_fore_idx)\n#test_loader = DataLoader(test_dataset,batch_size= batch ,shuffle=False)", "_____no_output_____" ], [ "n_batches = 3000//batch\nbg = []\nfor i in range(n_batches):\n torch.manual_seed(i)\n betag = torch.randn(3000,9)#torch.ones((250,9))/9\n bg.append( betag.requires_grad_() )", "_____no_output_____" ], [ "len(bg)", "_____no_output_____" ], [ "#H = np.zeros((27000,27000))\nfor i, data in enumerate(train_loader, 0):\n print(i) # only one batch\n inputs,_,_ = data\n inputs = torch.reshape(inputs,(27000,2))\n dis = torch.cdist(inputs,inputs)**2\n gamma = -1/torch.median(dis)\n print(gamma)\n H = torch.exp(gamma*dis)\n", "0\ntensor(-0.0948)\n" ], [ "H.shape", "_____no_output_____" ], [ "class Module2(nn.Module):\n def __init__(self):\n super(Module2, self).__init__()\n self.linear1 = nn.Linear(2,100)\n self.linear2 = nn.Linear(100,3)\n\n def forward(self,x):\n x = F.relu(self.linear1(x))\n x = self.linear2(x)\n return x", "_____no_output_____" ], [ "torch.manual_seed(1234)\nwhat_net = Module2().double()\n\n#what_net.load_state_dict(torch.load(\"type4_what_net.pt\"))\nwhat_net = what_net.to(\"cuda\")", "_____no_output_____" ], [ "def attn_avg(x,beta):\n y = torch.zeros([batch,2], dtype=torch.float64)\n y = y.to(\"cuda\")\n alpha = F.softmax(beta,dim=1) # alphas\n #print(alpha[0],x[0,:])\n for i in range(9): \n alpha1 = alpha[:,i] \n y = y + torch.mul(alpha1[:,None],x[:,i])\n return y,alpha\n", "_____no_output_____" ], [ "def calculate_attn_loss(dataloader,what,criter):\n what.eval()\n r_loss = 0\n alphas = []\n lbls = []\n pred = []\n fidices = []\n correct = 0\n tot = 0\n with torch.no_grad():\n for i, data in enumerate(dataloader, 0):\n inputs, labels,fidx= data\n lbls.append(labels)\n fidices.append(fidx)\n inputs = inputs.double()\n beta = bg[i] # beta for ith batch\n inputs, labels,beta = inputs.to(\"cuda\"),labels.to(\"cuda\"),beta.to(\"cuda\")\n avg,alpha = attn_avg(inputs,beta)\n alpha = alpha.to(\"cuda\")\n outputs = what(avg)\n _, predicted = torch.max(outputs.data, 1)\n correct += sum(predicted == labels)\n tot += len(predicted)\n pred.append(predicted.cpu().numpy())\n alphas.append(alpha.cpu().numpy())\n loss = criter(outputs, labels)\n r_loss += loss.item()\n alphas = np.concatenate(alphas,axis=0)\n pred = np.concatenate(pred,axis=0)\n lbls = np.concatenate(lbls,axis=0)\n fidices = np.concatenate(fidices,axis=0)\n #print(alphas.shape,pred.shape,lbls.shape,fidices.shape) \n analysis = analyse_data(alphas,lbls,pred,fidices)\n return r_loss/(i+1),analysis,correct.item(),tot,correct.item()/tot", "_____no_output_____" ], [ "# for param in what_net.parameters():\n# param.requires_grad = False", "_____no_output_____" ], [ "\ndef analyse_data(alphas,lbls,predicted,f_idx):\n '''\n analysis data is created here\n '''\n batch = len(predicted)\n amth,alth,ftpt,ffpt,ftpf,ffpf = 0,0,0,0,0,0\n for j in range (batch):\n focus = np.argmax(alphas[j])\n if(alphas[j][focus] >= 0.5):\n amth +=1\n else:\n alth +=1\n if(focus == f_idx[j] and predicted[j] == lbls[j]):\n ftpt += 1\n elif(focus != f_idx[j] and predicted[j] == lbls[j]):\n ffpt +=1\n elif(focus == f_idx[j] and predicted[j] != lbls[j]):\n ftpf +=1\n elif(focus != f_idx[j] and predicted[j] != lbls[j]):\n ffpf +=1\n #print(sum(predicted==lbls),ftpt+ffpt)\n return [ftpt,ffpt,ftpf,ffpf,amth,alth]", "_____no_output_____" ], [ "optim1 = []\nH= H.to(\"cpu\")\nfor i in range(n_batches):\n optim1.append(MyRmsprop([bg[i]],H=H,lr=0.1))\n #optim1.append(optim.RMSprop([bg[i]],lr=0.1))", "_____no_output_____" ], [ "# instantiate optimizer\noptimizer_what = optim.RMSprop(what_net.parameters(), lr=0.001)#, momentum=0.9)#,nesterov=True)\n\n\n \n\n\ncriterion = nn.CrossEntropyLoss()\nacti = []\nanalysis_data_tr = []\nanalysis_data_tst = []\nloss_curi_tr = []\nloss_curi_tst = []\nepochs = 200\n\n\n# calculate zeroth epoch loss and FTPT values\nrunning_loss,anlys_data,correct,total,accuracy = calculate_attn_loss(train_loader,what_net,criterion)\nprint('training epoch: [%d ] loss: %.3f correct: %.3f, total: %.3f, accuracy: %.3f' %(0,running_loss,correct,total,accuracy)) \nloss_curi_tr.append(running_loss)\nanalysis_data_tr.append(anlys_data)\n\n\n\n\n# training starts \nfor epoch in range(epochs): # loop over the dataset multiple times\n ep_lossi = []\n running_loss = 0.0\n what_net.train()\n for i, data in enumerate(train_loader, 0):\n # get the inputs\n inputs, labels,_ = data\n inputs = inputs.double()\n beta = bg[i] # alpha for ith batch\n #print(labels)\n inputs, labels,beta = inputs.to(\"cuda\"),labels.to(\"cuda\"),beta.to(\"cuda\")\n \n # zero the parameter gradients\n optimizer_what.zero_grad()\n optim1[i].zero_grad()\n \n # forward + backward + optimize\n avg,alpha = attn_avg(inputs,beta)\n outputs = what_net(avg) \n loss = criterion(outputs, labels)\n\n # print statistics\n running_loss += loss.item()\n #alpha.retain_grad()\n loss.backward(retain_graph=False)\n optimizer_what.step()\n optim1[i].step()\n\n\n running_loss_tr,anls_data,correct,total,accuracy = calculate_attn_loss(train_loader,what_net,criterion)\n analysis_data_tr.append(anls_data)\n loss_curi_tr.append(running_loss_tr) #loss per epoch\n print('training epoch: [%d ] loss: %.3f correct: %.3f, total: %.3f, accuracy: %.3f' %(epoch+1,running_loss_tr,correct,total,accuracy)) \n\n\n \n if running_loss_tr<=0.08:\n break\nprint('Finished Training run ')\nanalysis_data_tr = np.array(analysis_data_tr)\n\n", "training epoch: [0 ] loss: 1.478 correct: 971.000, total: 3000.000, accuracy: 0.324\ntraining epoch: [1 ] loss: 1.145 correct: 991.000, total: 3000.000, accuracy: 0.330\ntraining epoch: [2 ] loss: 1.138 correct: 1047.000, total: 3000.000, accuracy: 0.349\ntraining epoch: [3 ] loss: 1.124 correct: 1013.000, total: 3000.000, accuracy: 0.338\ntraining epoch: [4 ] loss: 1.103 correct: 1050.000, total: 3000.000, accuracy: 0.350\ntraining epoch: [5 ] loss: 1.097 correct: 1012.000, total: 3000.000, accuracy: 0.337\ntraining epoch: [6 ] loss: 1.093 correct: 1130.000, total: 3000.000, accuracy: 0.377\ntraining epoch: [7 ] loss: 1.091 correct: 1148.000, total: 3000.000, accuracy: 0.383\ntraining epoch: [8 ] loss: 1.089 correct: 1196.000, total: 3000.000, accuracy: 0.399\ntraining epoch: [9 ] loss: 1.086 correct: 1210.000, total: 3000.000, accuracy: 0.403\ntraining epoch: [10 ] loss: 1.083 correct: 1220.000, total: 3000.000, accuracy: 0.407\ntraining epoch: [11 ] loss: 1.080 correct: 1257.000, total: 3000.000, accuracy: 0.419\ntraining epoch: [12 ] loss: 1.076 correct: 1281.000, total: 3000.000, accuracy: 0.427\ntraining epoch: [13 ] loss: 1.070 correct: 1327.000, total: 3000.000, accuracy: 0.442\ntraining epoch: [14 ] loss: 1.063 correct: 1357.000, total: 3000.000, accuracy: 0.452\ntraining epoch: [15 ] loss: 1.055 correct: 1402.000, total: 3000.000, accuracy: 0.467\ntraining epoch: [16 ] loss: 1.045 correct: 1425.000, total: 3000.000, accuracy: 0.475\ntraining epoch: [17 ] loss: 1.032 correct: 1491.000, total: 3000.000, accuracy: 0.497\ntraining epoch: [18 ] loss: 1.018 correct: 1536.000, total: 3000.000, accuracy: 0.512\ntraining epoch: [19 ] loss: 1.003 correct: 1512.000, total: 3000.000, accuracy: 0.504\ntraining epoch: [20 ] loss: 0.989 correct: 1592.000, total: 3000.000, accuracy: 0.531\ntraining epoch: [21 ] loss: 0.986 correct: 1298.000, total: 3000.000, accuracy: 0.433\ntraining epoch: [22 ] loss: 0.995 correct: 1526.000, total: 3000.000, accuracy: 0.509\ntraining epoch: [23 ] loss: 0.979 correct: 1322.000, total: 3000.000, accuracy: 0.441\ntraining epoch: [24 ] loss: 0.946 correct: 1650.000, total: 3000.000, accuracy: 0.550\ntraining epoch: [25 ] loss: 0.904 correct: 1513.000, total: 3000.000, accuracy: 0.504\ntraining epoch: [26 ] loss: 0.875 correct: 1854.000, total: 3000.000, accuracy: 0.618\ntraining epoch: [27 ] loss: 0.850 correct: 1712.000, total: 3000.000, accuracy: 0.571\ntraining epoch: [28 ] loss: 0.828 correct: 1951.000, total: 3000.000, accuracy: 0.650\ntraining epoch: [29 ] loss: 0.807 correct: 1932.000, total: 3000.000, accuracy: 0.644\ntraining epoch: [30 ] loss: 0.788 correct: 2091.000, total: 3000.000, accuracy: 0.697\ntraining epoch: [31 ] loss: 0.770 correct: 2067.000, total: 3000.000, accuracy: 0.689\ntraining epoch: [32 ] loss: 0.752 correct: 2182.000, total: 3000.000, accuracy: 0.727\ntraining epoch: [33 ] loss: 0.736 correct: 2159.000, total: 3000.000, accuracy: 0.720\ntraining epoch: [34 ] loss: 0.721 correct: 2227.000, total: 3000.000, accuracy: 0.742\ntraining epoch: [35 ] loss: 0.707 correct: 2227.000, total: 3000.000, accuracy: 0.742\ntraining epoch: [36 ] loss: 0.694 correct: 2276.000, total: 3000.000, accuracy: 0.759\ntraining epoch: [37 ] loss: 0.682 correct: 2278.000, total: 3000.000, accuracy: 0.759\ntraining epoch: [38 ] loss: 0.670 correct: 2327.000, total: 3000.000, accuracy: 0.776\ntraining epoch: [39 ] loss: 0.660 correct: 2341.000, total: 3000.000, accuracy: 0.780\ntraining epoch: [40 ] loss: 0.650 correct: 2377.000, total: 3000.000, accuracy: 0.792\ntraining epoch: [41 ] loss: 0.640 correct: 2402.000, total: 3000.000, accuracy: 0.801\ntraining epoch: [42 ] loss: 0.631 correct: 2440.000, total: 3000.000, accuracy: 0.813\ntraining epoch: [43 ] loss: 0.623 correct: 2471.000, total: 3000.000, accuracy: 0.824\ntraining epoch: [44 ] loss: 0.615 correct: 2507.000, total: 3000.000, accuracy: 0.836\ntraining epoch: [45 ] loss: 0.607 correct: 2530.000, total: 3000.000, accuracy: 0.843\ntraining epoch: [46 ] loss: 0.600 correct: 2557.000, total: 3000.000, accuracy: 0.852\ntraining epoch: [47 ] loss: 0.592 correct: 2604.000, total: 3000.000, accuracy: 0.868\ntraining epoch: [48 ] loss: 0.583 correct: 2690.000, total: 3000.000, accuracy: 0.897\ntraining epoch: [49 ] loss: 0.576 correct: 2702.000, total: 3000.000, accuracy: 0.901\ntraining epoch: [50 ] loss: 0.570 correct: 2723.000, total: 3000.000, accuracy: 0.908\ntraining epoch: [51 ] loss: 0.564 correct: 2743.000, total: 3000.000, accuracy: 0.914\ntraining epoch: [52 ] loss: 0.558 correct: 2761.000, total: 3000.000, accuracy: 0.920\ntraining epoch: [53 ] loss: 0.552 correct: 2773.000, total: 3000.000, accuracy: 0.924\ntraining epoch: [54 ] loss: 0.547 correct: 2792.000, total: 3000.000, accuracy: 0.931\ntraining epoch: [55 ] loss: 0.542 correct: 2799.000, total: 3000.000, accuracy: 0.933\ntraining epoch: [56 ] loss: 0.537 correct: 2822.000, total: 3000.000, accuracy: 0.941\ntraining epoch: [57 ] loss: 0.533 correct: 2830.000, total: 3000.000, accuracy: 0.943\ntraining epoch: [58 ] loss: 0.528 correct: 2844.000, total: 3000.000, accuracy: 0.948\ntraining epoch: [59 ] loss: 0.524 correct: 2854.000, total: 3000.000, accuracy: 0.951\ntraining epoch: [60 ] loss: 0.519 correct: 2865.000, total: 3000.000, accuracy: 0.955\ntraining epoch: [61 ] loss: 0.515 correct: 2871.000, total: 3000.000, accuracy: 0.957\ntraining epoch: [62 ] loss: 0.511 correct: 2880.000, total: 3000.000, accuracy: 0.960\ntraining epoch: [63 ] loss: 0.507 correct: 2884.000, total: 3000.000, accuracy: 0.961\ntraining epoch: [64 ] loss: 0.503 correct: 2889.000, total: 3000.000, accuracy: 0.963\ntraining epoch: [65 ] loss: 0.499 correct: 2892.000, total: 3000.000, accuracy: 0.964\ntraining epoch: [66 ] loss: 0.496 correct: 2904.000, total: 3000.000, accuracy: 0.968\ntraining epoch: [67 ] loss: 0.492 correct: 2905.000, total: 3000.000, accuracy: 0.968\ntraining epoch: [68 ] loss: 0.488 correct: 2914.000, total: 3000.000, accuracy: 0.971\ntraining epoch: [69 ] loss: 0.485 correct: 2914.000, total: 3000.000, accuracy: 0.971\ntraining epoch: [70 ] loss: 0.482 correct: 2923.000, total: 3000.000, accuracy: 0.974\ntraining epoch: [71 ] loss: 0.478 correct: 2918.000, total: 3000.000, accuracy: 0.973\ntraining epoch: [72 ] loss: 0.475 correct: 2929.000, total: 3000.000, accuracy: 0.976\ntraining epoch: [73 ] loss: 0.472 correct: 2921.000, total: 3000.000, accuracy: 0.974\ntraining epoch: [74 ] loss: 0.469 correct: 2936.000, total: 3000.000, accuracy: 0.979\ntraining epoch: [75 ] loss: 0.465 correct: 2921.000, total: 3000.000, accuracy: 0.974\ntraining epoch: [76 ] loss: 0.462 correct: 2943.000, total: 3000.000, accuracy: 0.981\ntraining epoch: [77 ] loss: 0.459 correct: 2920.000, total: 3000.000, accuracy: 0.973\ntraining epoch: [78 ] loss: 0.456 correct: 2950.000, total: 3000.000, accuracy: 0.983\ntraining epoch: [79 ] loss: 0.454 correct: 2914.000, total: 3000.000, accuracy: 0.971\ntraining epoch: [80 ] loss: 0.451 correct: 2955.000, total: 3000.000, accuracy: 0.985\ntraining epoch: [81 ] loss: 0.448 correct: 2877.000, total: 3000.000, accuracy: 0.959\ntraining epoch: [82 ] loss: 0.446 correct: 2947.000, total: 3000.000, accuracy: 0.982\ntraining epoch: [83 ] loss: 0.445 correct: 2742.000, total: 3000.000, accuracy: 0.914\ntraining epoch: [84 ] loss: 0.445 correct: 2825.000, total: 3000.000, accuracy: 0.942\ntraining epoch: [85 ] loss: 0.445 correct: 2429.000, total: 3000.000, accuracy: 0.810\ntraining epoch: [86 ] loss: 0.446 correct: 2488.000, total: 3000.000, accuracy: 0.829\ntraining epoch: [87 ] loss: 0.447 correct: 2191.000, total: 3000.000, accuracy: 0.730\ntraining epoch: [88 ] loss: 0.444 correct: 2397.000, total: 3000.000, accuracy: 0.799\ntraining epoch: [89 ] loss: 0.440 correct: 2298.000, total: 3000.000, accuracy: 0.766\ntraining epoch: [90 ] loss: 0.434 correct: 2677.000, total: 3000.000, accuracy: 0.892\ntraining epoch: [91 ] loss: 0.429 correct: 2613.000, total: 3000.000, accuracy: 0.871\ntraining epoch: [92 ] loss: 0.425 correct: 2889.000, total: 3000.000, accuracy: 0.963\ntraining epoch: [93 ] loss: 0.421 correct: 2830.000, total: 3000.000, accuracy: 0.943\ntraining epoch: [94 ] loss: 0.418 correct: 2959.000, total: 3000.000, accuracy: 0.986\ntraining epoch: [95 ] loss: 0.415 correct: 2907.000, total: 3000.000, accuracy: 0.969\ntraining epoch: [96 ] loss: 0.412 correct: 2970.000, total: 3000.000, accuracy: 0.990\ntraining epoch: [97 ] loss: 0.410 correct: 2937.000, total: 3000.000, accuracy: 0.979\ntraining epoch: [98 ] loss: 0.407 correct: 2972.000, total: 3000.000, accuracy: 0.991\ntraining epoch: [99 ] loss: 0.405 correct: 2947.000, total: 3000.000, accuracy: 0.982\ntraining epoch: [100 ] loss: 0.403 correct: 2975.000, total: 3000.000, accuracy: 0.992\ntraining epoch: [101 ] loss: 0.401 correct: 2962.000, total: 3000.000, accuracy: 0.987\ntraining epoch: [102 ] loss: 0.399 correct: 2976.000, total: 3000.000, accuracy: 0.992\ntraining epoch: [103 ] loss: 0.396 correct: 2965.000, total: 3000.000, accuracy: 0.988\ntraining epoch: [104 ] loss: 0.394 correct: 2978.000, total: 3000.000, accuracy: 0.993\ntraining epoch: [105 ] loss: 0.392 correct: 2968.000, total: 3000.000, accuracy: 0.989\ntraining epoch: [106 ] loss: 0.390 correct: 2980.000, total: 3000.000, accuracy: 0.993\ntraining epoch: [107 ] loss: 0.388 correct: 2969.000, total: 3000.000, accuracy: 0.990\ntraining epoch: [108 ] loss: 0.386 correct: 2982.000, total: 3000.000, accuracy: 0.994\ntraining epoch: [109 ] loss: 0.384 correct: 2970.000, total: 3000.000, accuracy: 0.990\ntraining epoch: [110 ] loss: 0.382 correct: 2983.000, total: 3000.000, accuracy: 0.994\ntraining epoch: [111 ] loss: 0.380 correct: 2973.000, total: 3000.000, accuracy: 0.991\ntraining epoch: [112 ] loss: 0.378 correct: 2985.000, total: 3000.000, accuracy: 0.995\ntraining epoch: [113 ] loss: 0.376 correct: 2974.000, total: 3000.000, accuracy: 0.991\ntraining epoch: [114 ] loss: 0.375 correct: 2987.000, total: 3000.000, accuracy: 0.996\ntraining epoch: [115 ] loss: 0.373 correct: 2976.000, total: 3000.000, accuracy: 0.992\ntraining epoch: [116 ] loss: 0.371 correct: 2987.000, total: 3000.000, accuracy: 0.996\ntraining epoch: [117 ] loss: 0.369 correct: 2977.000, total: 3000.000, accuracy: 0.992\ntraining epoch: [118 ] loss: 0.367 correct: 2987.000, total: 3000.000, accuracy: 0.996\ntraining epoch: [119 ] loss: 0.365 correct: 2979.000, total: 3000.000, accuracy: 0.993\ntraining epoch: [120 ] loss: 0.363 correct: 2987.000, total: 3000.000, accuracy: 0.996\ntraining epoch: [121 ] loss: 0.361 correct: 2980.000, total: 3000.000, accuracy: 0.993\ntraining epoch: [122 ] loss: 0.360 correct: 2987.000, total: 3000.000, accuracy: 0.996\ntraining epoch: [123 ] loss: 0.358 correct: 2979.000, total: 3000.000, accuracy: 0.993\ntraining epoch: [124 ] loss: 0.356 correct: 2987.000, total: 3000.000, accuracy: 0.996\ntraining epoch: [125 ] loss: 0.354 correct: 2981.000, total: 3000.000, accuracy: 0.994\ntraining epoch: [126 ] loss: 0.352 correct: 2987.000, total: 3000.000, accuracy: 0.996\ntraining epoch: [127 ] loss: 0.351 correct: 2980.000, total: 3000.000, accuracy: 0.993\ntraining epoch: [128 ] loss: 0.349 correct: 2987.000, total: 3000.000, accuracy: 0.996\ntraining epoch: [129 ] loss: 0.347 correct: 2979.000, total: 3000.000, accuracy: 0.993\ntraining epoch: [130 ] loss: 0.345 correct: 2987.000, total: 3000.000, accuracy: 0.996\ntraining epoch: [131 ] loss: 0.344 correct: 2978.000, total: 3000.000, accuracy: 0.993\ntraining epoch: [132 ] loss: 0.342 correct: 2986.000, total: 3000.000, accuracy: 0.995\ntraining epoch: [133 ] loss: 0.341 correct: 2974.000, total: 3000.000, accuracy: 0.991\ntraining epoch: [134 ] loss: 0.339 correct: 2990.000, total: 3000.000, accuracy: 0.997\ntraining epoch: [135 ] loss: 0.338 correct: 2959.000, total: 3000.000, accuracy: 0.986\ntraining epoch: [136 ] loss: 0.337 correct: 2986.000, total: 3000.000, accuracy: 0.995\ntraining epoch: [137 ] loss: 0.336 correct: 2946.000, total: 3000.000, accuracy: 0.982\ntraining epoch: [138 ] loss: 0.335 correct: 2974.000, total: 3000.000, accuracy: 0.991\ntraining epoch: [139 ] loss: 0.335 correct: 2924.000, total: 3000.000, accuracy: 0.975\ntraining epoch: [140 ] loss: 0.335 correct: 2956.000, total: 3000.000, accuracy: 0.985\ntraining epoch: [141 ] loss: 0.335 correct: 2900.000, total: 3000.000, accuracy: 0.967\ntraining epoch: [142 ] loss: 0.333 correct: 2941.000, total: 3000.000, accuracy: 0.980\ntraining epoch: [143 ] loss: 0.332 correct: 2896.000, total: 3000.000, accuracy: 0.965\ntraining epoch: [144 ] loss: 0.329 correct: 2955.000, total: 3000.000, accuracy: 0.985\ntraining epoch: [145 ] loss: 0.327 correct: 2924.000, total: 3000.000, accuracy: 0.975\ntraining epoch: [146 ] loss: 0.324 correct: 2973.000, total: 3000.000, accuracy: 0.991\ntraining epoch: [147 ] loss: 0.322 correct: 2940.000, total: 3000.000, accuracy: 0.980\ntraining epoch: [148 ] loss: 0.319 correct: 2985.000, total: 3000.000, accuracy: 0.995\ntraining epoch: [149 ] loss: 0.317 correct: 2960.000, total: 3000.000, accuracy: 0.987\ntraining epoch: [150 ] loss: 0.315 correct: 2990.000, total: 3000.000, accuracy: 0.997\ntraining epoch: [151 ] loss: 0.313 correct: 2972.000, total: 3000.000, accuracy: 0.991\ntraining epoch: [152 ] loss: 0.311 correct: 2992.000, total: 3000.000, accuracy: 0.997\ntraining epoch: [153 ] loss: 0.309 correct: 2977.000, total: 3000.000, accuracy: 0.992\ntraining epoch: [154 ] loss: 0.308 correct: 2993.000, total: 3000.000, accuracy: 0.998\ntraining epoch: [155 ] loss: 0.306 correct: 2981.000, total: 3000.000, accuracy: 0.994\ntraining epoch: [156 ] loss: 0.305 correct: 2991.000, total: 3000.000, accuracy: 0.997\ntraining epoch: [157 ] loss: 0.303 correct: 2983.000, total: 3000.000, accuracy: 0.994\ntraining epoch: [158 ] loss: 0.301 correct: 2991.000, total: 3000.000, accuracy: 0.997\ntraining epoch: [159 ] loss: 0.300 correct: 2986.000, total: 3000.000, accuracy: 0.995\ntraining epoch: [160 ] loss: 0.299 correct: 2991.000, total: 3000.000, accuracy: 0.997\ntraining epoch: [161 ] loss: 0.297 correct: 2987.000, total: 3000.000, accuracy: 0.996\ntraining epoch: [162 ] loss: 0.296 correct: 2991.000, total: 3000.000, accuracy: 0.997\ntraining epoch: [163 ] loss: 0.294 correct: 2989.000, total: 3000.000, accuracy: 0.996\ntraining epoch: [164 ] loss: 0.293 correct: 2991.000, total: 3000.000, accuracy: 0.997\ntraining epoch: [165 ] loss: 0.291 correct: 2990.000, total: 3000.000, accuracy: 0.997\ntraining epoch: [166 ] loss: 0.290 correct: 2991.000, total: 3000.000, accuracy: 0.997\ntraining epoch: [167 ] loss: 0.289 correct: 2991.000, total: 3000.000, accuracy: 0.997\ntraining epoch: [168 ] loss: 0.287 correct: 2991.000, total: 3000.000, accuracy: 0.997\ntraining epoch: [169 ] loss: 0.286 correct: 2991.000, total: 3000.000, accuracy: 0.997\ntraining epoch: [170 ] loss: 0.285 correct: 2991.000, total: 3000.000, accuracy: 0.997\ntraining epoch: [171 ] loss: 0.283 correct: 2991.000, total: 3000.000, accuracy: 0.997\ntraining epoch: [172 ] loss: 0.282 correct: 2991.000, total: 3000.000, accuracy: 0.997\ntraining epoch: [173 ] loss: 0.281 correct: 2990.000, total: 3000.000, accuracy: 0.997\ntraining epoch: [174 ] loss: 0.279 correct: 2991.000, total: 3000.000, accuracy: 0.997\ntraining epoch: [175 ] loss: 0.278 correct: 2989.000, total: 3000.000, accuracy: 0.996\ntraining epoch: [176 ] loss: 0.277 correct: 2991.000, total: 3000.000, accuracy: 0.997\ntraining epoch: [177 ] loss: 0.275 correct: 2988.000, total: 3000.000, accuracy: 0.996\ntraining epoch: [178 ] loss: 0.274 correct: 2992.000, total: 3000.000, accuracy: 0.997\ntraining epoch: [179 ] loss: 0.273 correct: 2988.000, total: 3000.000, accuracy: 0.996\ntraining epoch: [180 ] loss: 0.272 correct: 2992.000, total: 3000.000, accuracy: 0.997\ntraining epoch: [181 ] loss: 0.270 correct: 2985.000, total: 3000.000, accuracy: 0.995\ntraining epoch: [182 ] loss: 0.269 correct: 2995.000, total: 3000.000, accuracy: 0.998\ntraining epoch: [183 ] loss: 0.268 correct: 2982.000, total: 3000.000, accuracy: 0.994\ntraining epoch: [184 ] loss: 0.267 correct: 2996.000, total: 3000.000, accuracy: 0.999\ntraining epoch: [185 ] loss: 0.266 correct: 2979.000, total: 3000.000, accuracy: 0.993\ntraining epoch: [186 ] loss: 0.265 correct: 2995.000, total: 3000.000, accuracy: 0.998\ntraining epoch: [187 ] loss: 0.264 correct: 2977.000, total: 3000.000, accuracy: 0.992\ntraining epoch: [188 ] loss: 0.264 correct: 2991.000, total: 3000.000, accuracy: 0.997\ntraining epoch: [189 ] loss: 0.263 correct: 2965.000, total: 3000.000, accuracy: 0.988\ntraining epoch: [190 ] loss: 0.262 correct: 2988.000, total: 3000.000, accuracy: 0.996\ntraining epoch: [191 ] loss: 0.262 correct: 2962.000, total: 3000.000, accuracy: 0.987\ntraining epoch: [192 ] loss: 0.261 correct: 2987.000, total: 3000.000, accuracy: 0.996\ntraining epoch: [193 ] loss: 0.260 correct: 2962.000, total: 3000.000, accuracy: 0.987\ntraining epoch: [194 ] loss: 0.258 correct: 2988.000, total: 3000.000, accuracy: 0.996\ntraining epoch: [195 ] loss: 0.257 correct: 2963.000, total: 3000.000, accuracy: 0.988\ntraining epoch: [196 ] loss: 0.255 correct: 2989.000, total: 3000.000, accuracy: 0.996\ntraining epoch: [197 ] loss: 0.254 correct: 2967.000, total: 3000.000, accuracy: 0.989\ntraining epoch: [198 ] loss: 0.252 correct: 2991.000, total: 3000.000, accuracy: 0.997\ntraining epoch: [199 ] loss: 0.250 correct: 2979.000, total: 3000.000, accuracy: 0.993\ntraining epoch: [200 ] loss: 0.249 correct: 2996.000, total: 3000.000, accuracy: 0.999\nFinished Training run \n" ], [ "columns = [\"epochs\", \"argmax > 0.5\" ,\"argmax < 0.5\", \"focus_true_pred_true\", \"focus_false_pred_true\", \"focus_true_pred_false\", \"focus_false_pred_false\" ]\ndf_train = pd.DataFrame()\ndf_test = pd.DataFrame()\ndf_train[columns[0]] = np.arange(0,epoch+2)\ndf_train[columns[1]] = analysis_data_tr[:,-2]/30\ndf_train[columns[2]] = analysis_data_tr[:,-1]/30\ndf_train[columns[3]] = analysis_data_tr[:,0]/30\ndf_train[columns[4]] = analysis_data_tr[:,1]/30\ndf_train[columns[5]] = analysis_data_tr[:,2]/30\ndf_train[columns[6]] = analysis_data_tr[:,3]/30", "_____no_output_____" ], [ "df_train", "_____no_output_____" ], [ "fig= plt.figure(figsize=(6,6))\nplt.plot(df_train[columns[0]],df_train[columns[3]], label =\"focus_true_pred_true \")\nplt.plot(df_train[columns[0]],df_train[columns[4]], label =\"focus_false_pred_true \")\nplt.plot(df_train[columns[0]],df_train[columns[5]], label =\"focus_true_pred_false \")\nplt.plot(df_train[columns[0]],df_train[columns[6]], label =\"focus_false_pred_false \")\nplt.title(\"On Train set\")\nplt.legend(loc='center left', bbox_to_anchor=(1, 0.5))\nplt.xlabel(\"epochs\")\nplt.ylabel(\"percentage of data\")\nplt.xticks([0,50,100,150,200])\n#plt.vlines(vline_list,min(min(df_train[columns[3]]/300),min(df_train[columns[4]]/300),min(df_train[columns[5]]/300),min(df_train[columns[6]]/300)), max(max(df_train[columns[3]]/300),max(df_train[columns[4]]/300),max(df_train[columns[5]]/300),max(df_train[columns[6]]/300)),linestyles='dotted')\nplt.show()\nfig.savefig(\"train_analysis.pdf\")\nfig.savefig(\"train_analysis.png\")", "_____no_output_____" ], [ "aph = []\nfor i in bg:\n aph.append(F.softmax(i,dim=1).detach().numpy())\naph = np.concatenate(aph,axis=0)\n# torch.save({\n# 'epoch': 500,\n# 'model_state_dict': what_net.state_dict(),\n# #'optimizer_state_dict': optimizer_what.state_dict(),\n# \"optimizer_alpha\":optim1,\n# \"FTPT_analysis\":analysis_data_tr,\n# \"alpha\":aph\n\n# }, \"type4_what_net_500.pt\")", "_____no_output_____" ], [ "aph[0]", "_____no_output_____" ], [ "xx,yy= np.meshgrid(np.arange(1,8,0.01),np.arange(2,9,0.01))\nX = np.concatenate((xx.reshape(-1,1),yy.reshape(-1,1)),axis=1)\nX = torch.Tensor(X).double().to(\"cuda\")\nY1 = what_net(X)", "_____no_output_____" ], [ "Y1 = Y1.to(\"cpu\")\nY1 = Y1.detach().numpy()\nY1 = torch.softmax(torch.Tensor(Y1),dim=1)\n_,Z4= torch.max(Y1,1)\nZ1 = Y1[:,0]\nZ2 = Y1[:,1]\nZ3 = Y1[:,2]", "_____no_output_____" ], [ "X = X.to(\"cpu\")", "_____no_output_____" ], [ "data = np.load(\"type_4_data.npy\",allow_pickle=True)\nx = data[0][\"X\"]\ny = data[0][\"Y\"]\nidx= []\nfor i in range(10):\n print(i,sum(y==i))\n idx.append(y==i)", "0 482\n1 485\n2 536\n3 504\n4 493\n5 513\n6 497\n7 486\n8 522\n9 482\n" ], [ "avrg = []\nwith torch.no_grad():\n for i, data in enumerate(train_loader):\n inputs , labels , fore_idx = data\n inputs = inputs.double()\n inputs, labels = inputs.to(\"cuda\"),labels.to(\"cuda\")\n beta = bg[i]\n beta = beta.to(\"cuda\")\n avg,alpha = attn_avg(inputs,beta)\n \n avrg.append(avg.detach().cpu().numpy())\navrg= np.concatenate(avrg,axis=0)", "_____no_output_____" ], [ "fig = plt.figure(figsize=(6,6))\n#plt.scatter(X[:,0],X[:,1],c=Z4)\nZ4 = Z4.reshape(xx.shape)\nplt.contourf(xx, yy, Z4, alpha=0.4)\nfor i in range(3):\n plt.scatter(x[idx[i],0],x[idx[i],1],label=\"class_\"+str(i),alpha=0.8)\nplt.legend(loc='center left', bbox_to_anchor=(1, 0.5))\nplt.scatter(avrg[:,0],avrg[:,1],alpha=0.2)\nplt.savefig(\"decision_boundary.png\",bbox_inches=\"tight\")\n", "_____no_output_____" ], [ "", "_____no_output_____" ] ] ]

[ "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" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "# Creating your own dataset from Google Images\n\n*by: Francisco Ingham and Jeremy Howard. Inspired by [Adrian Rosebrock](https://www.pyimagesearch.com/2017/12/04/how-to-create-a-deep-learning-dataset-using-google-images/)*", "_____no_output_____" ], [ "In this tutorial we will see how to easily create an image dataset through Google Images. **Note**: You will have to repeat these steps for any new category you want to Google (e.g once for dogs and once for cats).", "_____no_output_____" ] ], [ [ "from fastai.vision import *", "_____no_output_____" ] ], [ [ "## Get a list of URLs", "_____no_output_____" ], [ "### Search and scroll", "_____no_output_____" ], [ "Go to [Google Images](http://images.google.com) and search for the images you are interested in. The more specific you are in your Google Search, the better the results and the less manual pruning you will have to do.\n\nScroll down until you've seen all the images you want to download, or until you see a button that says 'Show more results'. All the images you scrolled past are now available to download. To get more, click on the button, and continue scrolling. The maximum number of images Google Images shows is 700.\n\nIt is a good idea to put things you want to exclude into the search query, for instance if you are searching for the Eurasian wolf, \"canis lupus lupus\", it might be a good idea to exclude other variants:\n\n \"canis lupus lupus\" -dog -arctos -familiaris -baileyi -occidentalis\n\nYou can also limit your results to show only photos by clicking on Tools and selecting Photos from the Type dropdown.", "_____no_output_____" ], [ "### Download into file", "_____no_output_____" ], [ "Now you must run some Javascript code in your browser which will save the URLs of all the images you want for you dataset.\n\nPress <kbd>Ctrl</kbd><kbd>Shift</kbd><kbd>J</kbd> in Windows/Linux and <kbd>Cmd</kbd><kbd>Opt</kbd><kbd>J</kbd> in Mac, and a small window the javascript 'Console' will appear. That is where you will paste the JavaScript commands.\n\nYou will need to get the urls of each of the images. Before running the following commands, you may want to disable ad blocking extensions (uBlock, AdBlockPlus etc.) in Chrome. Otherwise window.open() coomand doesn't work. Then you can run the following commands:\n\n```javascript\nurls = Array.from(document.querySelectorAll('.rg_di .rg_meta')).map(el=>JSON.parse(el.textContent).ou);\nwindow.open('data:text/csv;charset=utf-8,' + escape(urls.join('\\n')));\n```", "_____no_output_____" ], [ "### Create directory and upload urls file into your server", "_____no_output_____" ], [ "Choose an appropriate name for your labeled images. You can run these steps multiple times to create different labels.", "_____no_output_____" ] ], [ [ "folder = 'black'\nfile = 'urls_black.csv'", "_____no_output_____" ], [ "folder = 'teddys'\nfile = 'urls_teddys.csv'", "_____no_output_____" ], [ "folder = 'grizzly'\nfile = 'urls_grizzly.csv'", "_____no_output_____" ] ], [ [ "You will need to run this cell once per each category.", "_____no_output_____" ] ], [ [ "path = Path('data/bears')\ndest = path/folder\ndest.mkdir(parents=True, exist_ok=True)", "_____no_output_____" ], [ "path.ls()", "_____no_output_____" ] ], [ [ "Finally, upload your urls file. You just need to press 'Upload' in your working directory and select your file, then click 'Upload' for each of the displayed files.\n\n", "_____no_output_____" ], [ "## Download images", "_____no_output_____" ], [ "Now you will need to download your images from their respective urls.\n\nfast.ai has a function that allows you to do just that. You just have to specify the urls filename as well as the destination folder and this function will download and save all images that can be opened. If they have some problem in being opened, they will not be saved.\n\nLet's download our images! Notice you can choose a maximum number of images to be downloaded. In this case we will not download all the urls.\n\nYou will need to run this line once for every category.", "_____no_output_____" ] ], [ [ "classes = ['teddys','grizzly','black']", "_____no_output_____" ], [ "download_images(path/file, dest, max_pics=200)", "_____no_output_____" ], [ "# If you have problems download, try with `max_workers=0` to see exceptions:\ndownload_images(path/file, dest, max_pics=20, max_workers=0)", "_____no_output_____" ] ], [ [ "Then we can remove any images that can't be opened:", "_____no_output_____" ] ], [ [ "for c in classes:\n print(c)\n verify_images(path/c, delete=True, max_size=500)", "teddys\n" ] ], [ [ "## View data", "_____no_output_____" ] ], [ [ "np.random.seed(42)\ndata = ImageDataBunch.from_folder(path, train=\".\", valid_pct=0.2,\n ds_tfms=get_transforms(), size=224, num_workers=4).normalize(imagenet_stats)", "_____no_output_____" ], [ "# If you already cleaned your data, run this cell instead of the one before\n# np.random.seed(42)\n# data = ImageDataBunch.from_csv(path, folder=\".\", valid_pct=0.2, csv_labels='cleaned.csv',\n# ds_tfms=get_transforms(), size=224, num_workers=4).normalize(imagenet_stats)", "_____no_output_____" ] ], [ [ "Good! Let's take a look at some of our pictures then.", "_____no_output_____" ] ], [ [ "data.classes", "_____no_output_____" ], [ "data.show_batch(rows=3, figsize=(7,8))", "_____no_output_____" ], [ "data.classes, data.c, len(data.train_ds), len(data.valid_ds)", "_____no_output_____" ] ], [ [ "## Train model", "_____no_output_____" ] ], [ [ "learn = cnn_learner(data, models.resnet34, metrics=error_rate)", "_____no_output_____" ], [ "learn.fit_one_cycle(4)", "_____no_output_____" ], [ "learn.save('stage-1')", "_____no_output_____" ], [ "learn.unfreeze()", "_____no_output_____" ], [ "learn.lr_find()", "_____no_output_____" ], [ "learn.recorder.plot()", "_____no_output_____" ], [ "learn.fit_one_cycle(2, max_lr=slice(3e-5,3e-4))", "_____no_output_____" ], [ "learn.save('stage-2')", "_____no_output_____" ] ], [ [ "## Interpretation", "_____no_output_____" ] ], [ [ "learn.load('stage-2');", "_____no_output_____" ], [ "interp = ClassificationInterpretation.from_learner(learn)", "_____no_output_____" ], [ "interp.plot_confusion_matrix()", "_____no_output_____" ] ], [ [ "## Cleaning Up\n\nSome of our top losses aren't due to bad performance by our model. There are images in our data set that shouldn't be.\n\nUsing the `ImageCleaner` widget from `fastai.widgets` we can prune our top losses, removing photos that don't belong.", "_____no_output_____" ] ], [ [ "from fastai.widgets import *", "_____no_output_____" ] ], [ [ "First we need to get the file paths from our top_losses. We can do this with `.from_toplosses`. We then feed the top losses indexes and corresponding dataset to `ImageCleaner`.\n\nNotice that the widget will not delete images directly from disk but it will create a new csv file `cleaned.csv` from where you can create a new ImageDataBunch with the corrected labels to continue training your model.", "_____no_output_____" ], [ "In order to clean the entire set of images, we need to create a new dataset without the split. The video lecture demostrated the use of the `ds_type` param which no longer has any effect. See [the thread](https://forums.fast.ai/t/duplicate-widget/30975/10) for more details.", "_____no_output_____" ] ], [ [ "db = (ImageList.from_folder(path)\n .no_split()\n .label_from_folder()\n .transform(get_transforms(), size=224)\n .databunch()\n )", "_____no_output_____" ], [ "# If you already cleaned your data using indexes from `from_toplosses`,\n# run this cell instead of the one before to proceed with removing duplicates.\n# Otherwise all the results of the previous step would be overwritten by\n# the new run of `ImageCleaner`.\n\n# db = (ImageList.from_csv(path, 'cleaned.csv', folder='.')\n# .no_split()\n# .label_from_df()\n# .transform(get_transforms(), size=224)\n# .databunch()\n# )", "_____no_output_____" ] ], [ [ "Then we create a new learner to use our new databunch with all the images.", "_____no_output_____" ] ], [ [ "learn_cln = cnn_learner(db, models.resnet34, metrics=error_rate)\n\nlearn_cln.load('stage-2');", "_____no_output_____" ], [ "ds, idxs = DatasetFormatter().from_toplosses(learn_cln)", "_____no_output_____" ] ], [ [ "Make sure you're running this notebook in Jupyter Notebook, not Jupyter Lab. That is accessible via [/tree](/tree), not [/lab](/lab). Running the `ImageCleaner` widget in Jupyter Lab is [not currently supported](https://github.com/fastai/fastai/issues/1539).", "_____no_output_____" ] ], [ [ "ImageCleaner(ds, idxs, path)", "_____no_output_____" ] ], [ [ "Flag photos for deletion by clicking 'Delete'. Then click 'Next Batch' to delete flagged photos and keep the rest in that row. `ImageCleaner` will show you a new row of images until there are no more to show. In this case, the widget will show you images until there are none left from `top_losses.ImageCleaner(ds, idxs)`", "_____no_output_____" ], [ "You can also find duplicates in your dataset and delete them! To do this, you need to run `.from_similars` to get the potential duplicates' ids and then run `ImageCleaner` with `duplicates=True`. The API works in a similar way as with misclassified images: just choose the ones you want to delete and click 'Next Batch' until there are no more images left.", "_____no_output_____" ], [ "Make sure to recreate the databunch and `learn_cln` from the `cleaned.csv` file. Otherwise the file would be overwritten from scratch, loosing all the results from cleaning the data from toplosses.", "_____no_output_____" ] ], [ [ "ds, idxs = DatasetFormatter().from_similars(learn_cln)", "Getting activations...\n" ], [ "ImageCleaner(ds, idxs, path, duplicates=True)", "_____no_output_____" ] ], [ [ "Remember to recreate your ImageDataBunch from your `cleaned.csv` to include the changes you made in your data!", "_____no_output_____" ], [ "## Putting your model in production", "_____no_output_____" ], [ "First thing first, let's export the content of our `Learner` object for production:", "_____no_output_____" ] ], [ [ "learn.export()", "_____no_output_____" ] ], [ [ "This will create a file named 'export.pkl' in the directory where we were working that contains everything we need to deploy our model (the model, the weights but also some metadata like the classes or the transforms/normalization used).", "_____no_output_____" ], [ "You probably want to use CPU for inference, except at massive scale (and you almost certainly don't need to train in real-time). If you don't have a GPU that happens automatically. You can test your model on CPU like so:", "_____no_output_____" ] ], [ [ "defaults.device = torch.device('cpu')", "_____no_output_____" ], [ "img = open_image(path/'black'/'00000021.jpg')\nimg", "_____no_output_____" ] ], [ [ "We create our `Learner` in production enviromnent like this, jsut make sure that `path` contains the file 'export.pkl' from before.", "_____no_output_____" ] ], [ [ "learn = load_learner(path)", "_____no_output_____" ], [ "pred_class,pred_idx,outputs = learn.predict(img)\npred_class", "_____no_output_____" ] ], [ [ "So you might create a route something like this ([thanks](https://github.com/simonw/cougar-or-not) to Simon Willison for the structure of this code):\n\n```python\[email protected](\"/classify-url\", methods=[\"GET\"])\nasync def classify_url(request):\n bytes = await get_bytes(request.query_params[\"url\"])\n img = open_image(BytesIO(bytes))\n _,_,losses = learner.predict(img)\n return JSONResponse({\n \"predictions\": sorted(\n zip(cat_learner.data.classes, map(float, losses)),\n key=lambda p: p[1],\n reverse=True\n )\n })\n```\n\n(This example is for the [Starlette](https://www.starlette.io/) web app toolkit.)", "_____no_output_____" ], [ "## Things that can go wrong", "_____no_output_____" ], [ "- Most of the time things will train fine with the defaults\n- There's not much you really need to tune (despite what you've heard!)\n- Most likely are\n - Learning rate\n - Number of epochs", "_____no_output_____" ], [ "### Learning rate (LR) too high", "_____no_output_____" ] ], [ [ "learn = cnn_learner(data, models.resnet34, metrics=error_rate)", "_____no_output_____" ], [ "learn.fit_one_cycle(1, max_lr=0.5)", "Total time: 00:13\nepoch train_loss valid_loss error_rate \n1 12.220007 1144188288.000000 0.765957 (00:13)\n\n" ] ], [ [ "### Learning rate (LR) too low", "_____no_output_____" ] ], [ [ "learn = cnn_learner(data, models.resnet34, metrics=error_rate)", "_____no_output_____" ] ], [ [ "Previously we had this result:\n\n```\nTotal time: 00:57\nepoch train_loss valid_loss error_rate\n1 1.030236 0.179226 0.028369 (00:14)\n2 0.561508 0.055464 0.014184 (00:13)\n3 0.396103 0.053801 0.014184 (00:13)\n4 0.316883 0.050197 0.021277 (00:15)\n```", "_____no_output_____" ] ], [ [ "learn.fit_one_cycle(5, max_lr=1e-5)", "Total time: 01:07\nepoch train_loss valid_loss error_rate\n1 1.349151 1.062807 0.609929 (00:13)\n2 1.373262 1.045115 0.546099 (00:13)\n3 1.346169 1.006288 0.468085 (00:13)\n4 1.334486 0.978713 0.453901 (00:13)\n5 1.320978 0.978108 0.446809 (00:13)\n\n" ], [ "learn.recorder.plot_losses()", "_____no_output_____" ] ], [ [ "As well as taking a really long time, it's getting too many looks at each image, so may overfit.", "_____no_output_____" ], [ "### Too few epochs", "_____no_output_____" ] ], [ [ "learn = cnn_learner(data, models.resnet34, metrics=error_rate, pretrained=False)", "_____no_output_____" ], [ "learn.fit_one_cycle(1)", "Total time: 00:14\nepoch train_loss valid_loss error_rate\n1 0.602823 0.119616 0.049645 (00:14)\n\n" ] ], [ [ "### Too many epochs", "_____no_output_____" ] ], [ [ "np.random.seed(42)\ndata = ImageDataBunch.from_folder(path, train=\".\", valid_pct=0.9, bs=32, \n ds_tfms=get_transforms(do_flip=False, max_rotate=0, max_zoom=1, max_lighting=0, max_warp=0\n ),size=224, num_workers=4).normalize(imagenet_stats)", "_____no_output_____" ], [ "learn = cnn_learner(data, models.resnet50, metrics=error_rate, ps=0, wd=0)\nlearn.unfreeze()", "_____no_output_____" ], [ "learn.fit_one_cycle(40, slice(1e-6,1e-4))", "Total time: 06:39\nepoch train_loss valid_loss error_rate\n1 1.513021 1.041628 0.507326 (00:13)\n2 1.290093 0.994758 0.443223 (00:09)\n3 1.185764 0.936145 0.410256 (00:09)\n4 1.117229 0.838402 0.322344 (00:09)\n5 1.022635 0.734872 0.252747 (00:09)\n6 0.951374 0.627288 0.192308 (00:10)\n7 0.916111 0.558621 0.184982 (00:09)\n8 0.839068 0.503755 0.177656 (00:09)\n9 0.749610 0.433475 0.144689 (00:09)\n10 0.678583 0.367560 0.124542 (00:09)\n11 0.615280 0.327029 0.100733 (00:10)\n12 0.558776 0.298989 0.095238 (00:09)\n13 0.518109 0.266998 0.084249 (00:09)\n14 0.476290 0.257858 0.084249 (00:09)\n15 0.436865 0.227299 0.067766 (00:09)\n16 0.457189 0.236593 0.078755 (00:10)\n17 0.420905 0.240185 0.080586 (00:10)\n18 0.395686 0.255465 0.082418 (00:09)\n19 0.373232 0.263469 0.080586 (00:09)\n20 0.348988 0.258300 0.080586 (00:10)\n21 0.324616 0.261346 0.080586 (00:09)\n22 0.311310 0.236431 0.071429 (00:09)\n23 0.328342 0.245841 0.069597 (00:10)\n24 0.306411 0.235111 0.064103 (00:10)\n25 0.289134 0.227465 0.069597 (00:09)\n26 0.284814 0.226022 0.064103 (00:09)\n27 0.268398 0.222791 0.067766 (00:09)\n28 0.255431 0.227751 0.073260 (00:10)\n29 0.240742 0.235949 0.071429 (00:09)\n30 0.227140 0.225221 0.075092 (00:09)\n31 0.213877 0.214789 0.069597 (00:09)\n32 0.201631 0.209382 0.062271 (00:10)\n33 0.189988 0.210684 0.065934 (00:09)\n34 0.181293 0.214666 0.073260 (00:09)\n35 0.184095 0.222575 0.073260 (00:09)\n36 0.194615 0.229198 0.076923 (00:10)\n37 0.186165 0.218206 0.075092 (00:09)\n38 0.176623 0.207198 0.062271 (00:10)\n39 0.166854 0.207256 0.065934 (00:10)\n40 0.162692 0.206044 0.062271 (00:09)\n\n" ] ] ]

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

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import os\nimport pandas as pd\nfrom xml.dom import minidom\nimport numpy as np\nfrom mpl_toolkits import mplot3d\nimport matplotlib.pyplot as plt\nimport seaborn as sns\nfrom sklearn.metrics import confusion_matrix, f1_score, accuracy_score, precision_score, recall_score, matthews_corrcoef, r2_score, roc_auc_score, auc, roc_curve", "_____no_output_____" ], [ "def load_semeval(t = 'train', lang = 'en'):\n semeval_keys = {\n 'correct': 2,\n 'incorrect': 1,\n 'contradictory': 0\n }\n \n file = minidom.parse('../../datasets/semeval2013-3way-' + lang + '/' + t + '.xml')\n \n for exercise in file.getElementsByTagName('exercise'):\n \n yield (\n [ reference.firstChild.data for reference in exercise.getElementsByTagName('reference') ],\n [ (answer.firstChild.data, answer.attributes['accuracy'].value) for answer in exercise.getElementsByTagName('answer') ]\n )", "_____no_output_____" ], [ "semeval = pd.DataFrame([\n (t, len(r), len(a))\n for t in [ 'train', 'unseen_answers', 'unseen_questions', 'unseen_domains' ]\n for r, a in load_semeval(t, 'en') \n], columns = [ 'dataset', 'references', 'answers' ])", "_____no_output_____" ], [ "plt.axis('equal');\nplt.pie(semeval.groupby([ 'dataset' ]).sum().reset_index()['answers'], \n labels = semeval.groupby([ 'dataset' ]).count().reset_index()['dataset'].map(str) + \n \" (\" + \n semeval.groupby([ 'dataset' ]).sum().reset_index()['answers'].map(str) +\n \")\"\n)\nplt.title('Answer Count')\nplt.show()", "_____no_output_____" ], [ "plt.axis('equal');\nplt.pie(semeval.groupby([ 'dataset' ]).sum().reset_index()['references'], \n labels = semeval.groupby([ 'dataset' ]).count().reset_index()['dataset'].map(str) + \n \" (\" + \n semeval.groupby([ 'dataset' ]).sum().reset_index()['references'].map(str) +\n \")\"\n)\nplt.title('Reference Count')\nplt.show()", "_____no_output_____" ], [ "pds = semeval.groupby([ 'dataset' ]).count().reset_index()\npds['references'] = semeval.groupby([ 'dataset' ]).sum().reset_index()['references'] / semeval.groupby([ 'dataset' ]).count().reset_index()['references']\n\nplt.figure(figsize=(8,4))\nsns.barplot(x = 'dataset', y = 'references', data=pds)\nplt.title('Avg. Refences Per Exercise')\nplt.show()", "_____no_output_____" ], [ "sns.barplot(\n x = 'references',\n y = 'answers',\n hue = 'dataset',\n ci= None,\n data = semeval\n)\nplt.title('Distribution of Answer Count to Reference Count')\nplt.show()", "_____no_output_____" ], [ "sns.scatterplot(\n x = 'references',\n y = 'answers',\n hue = 'dataset',\n s = 50,\n data = semeval[(semeval['dataset'] == 'unseen_questions') | (semeval['dataset'] == 'unseen_domains')]\n)\nplt.legend(loc='lower right')\nplt.title('Distribution of Answer Count to Reference Count')\nplt.show()", "_____no_output_____" ], [ "df = pd.DataFrame([\n (t, len(r), ans[1], 1)\n for t in [ 'train', 'unseen_answers', 'unseen_questions', 'unseen_domains' ]\n for r, a in load_semeval(t, 'en')\n for ans in a\n], columns = [ 'dataset', 'references', 'label', 'labels' ]).groupby([ 'dataset', 'label', 'references' ]).count().reset_index()\n\nfor t in [ 'train', 'unseen_answers', 'unseen_questions', 'unseen_domains' ]:\n sns.barplot(\n x = 'references',\n y = 'labels',\n hue = 'label',\n ci = None,\n data = df[df['dataset'] == t]\n )\n plt.title('Distribution of Answer Count to Reference Count (%s)' % t)\n plt.show()", "_____no_output_____" ] ] ]

[ "code" ]

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

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import xarray as xa\n", "/Users/bell/anaconda2/envs/py37/lib/python3.7/site-packages/dask/config.py:168: YAMLLoadWarning: calling yaml.load() without Loader=... is deprecated, as the default Loader is unsafe. Please read https://msg.pyyaml.org/load for full details.\n data = yaml.load(f.read()) or {}\n/Users/bell/anaconda2/envs/py37/lib/python3.7/site-packages/distributed/config.py:20: YAMLLoadWarning: calling yaml.load() without Loader=... is deprecated, as the default Loader is unsafe. Please read https://msg.pyyaml.org/load for full details.\n defaults = yaml.load(f)\n" ], [ "%matplotlib inline\nimport matplotlib.pyplot as plt\nimport numpy as np\n\nimport cartopy\nimport cartopy.crs as ccrs\nfrom cartopy.feature import NaturalEarthFeature\n\n\ndef sample_data(shape=(20, 30)):\n \"\"\"\n Returns ``(x, y, u, v, crs)`` of some vector data\n computed mathematically. The returned crs will be a rotated\n pole CRS, meaning that the vectors will be unevenly spaced in\n regular PlateCarree space.\n\n \"\"\"\n crs = ccrs.RotatedPole(pole_longitude=177.5, pole_latitude=37.5)\n\n x = np.linspace(311.9, 391.1, shape[1])\n y = np.linspace(-23.6, 24.8, shape[0])\n\n x2d, y2d = np.meshgrid(x, y)\n u = 10 * (2 * np.cos(2 * np.deg2rad(x2d) + 3 * np.deg2rad(y2d + 30)) ** 2)\n v = 20 * np.cos(6 * np.deg2rad(x2d))\n\n return x, y, u, v, crs", "_____no_output_____" ], [ "xdf = xa.open_dataset('/Users/bell/Downloads/for_Phyllis/uo_168_180_10_mon2014.nc')\nxdf.info", "_____no_output_____" ], [ "u = xdf.uo.values[1,1,:,:]\nu[u>999]=np.nan\n\nv = xdf.vo.values[1,1,:,:]\nv[v>999]=np.nan", "/Users/bell/anaconda2/envs/py37/lib/python3.7/site-packages/ipykernel_launcher.py:2: RuntimeWarning: invalid value encountered in greater\n \n/Users/bell/anaconda2/envs/py37/lib/python3.7/site-packages/ipykernel_launcher.py:5: RuntimeWarning: invalid value encountered in greater\n \"\"\"\n" ], [ "fig = plt.figure(1,figsize=(9,9))\nax = plt.axes(projection=ccrs.Orthographic(-135, 90))\n\ncoast = NaturalEarthFeature(category='physical', scale='50m',\n facecolor='grey', name='land')\n\nax.set_global()\nax.gridlines()\n\nax.quiver(xdf.lon.values, xdf.lat.values, u, v, transform=ccrs.PlateCarree())\nax.set_extent([-180, -130, 55, 90], crs=ccrs.PlateCarree())\nfeature = ax.add_feature(coast, edgecolor='black')\n\nfig.savefig('demo.png',dpi=300)", "/Users/bell/anaconda2/envs/py37/lib/python3.7/site-packages/cartopy/mpl/geoaxes.py:1752: RuntimeWarning: invalid value encountered in less\n u, v = self.projection.transform_vectors(t, x, y, u, v)\n/Users/bell/anaconda2/envs/py37/lib/python3.7/site-packages/cartopy/mpl/geoaxes.py:1752: RuntimeWarning: invalid value encountered in greater\n u, v = self.projection.transform_vectors(t, x, y, u, v)\n" ] ] ]

[ "code" ]

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

[ "ADSL" ]

[ "ADSL" ]

[ "ADSL" ]

[ [ [ "## Observations and Insights ", "_____no_output_____" ] ], [ [ "# Dependencies and Setup\nimport matplotlib.pyplot as plt\nimport pandas as pd\nimport scipy.stats as stats\nimport numpy as np\nfrom scipy.stats import linregress\n\n# Study data files\nmouse_metadata_path = \"data/Mouse_metadata.csv\"\nstudy_results_path = \"data/Study_results.csv\"\n\n# Read the mouse data and the study results\nmouse_metadata = pd.read_csv(mouse_metadata_path)\nstudy_results = pd.read_csv(study_results_path)\n\n# Combine the data into a single dataset\nstudy_data_complete = pd.merge(study_results, mouse_metadata, how=\"left\", on=\"Mouse ID\")\n\n# Display the data table for preview\nstudy_data_complete.head()", "_____no_output_____" ], [ "# Checking the number of mice.\nstudy_data_complete['Mouse ID'].nunique()", "_____no_output_____" ], [ "# Optional: Get all the data for the duplicate mouse ID. \nstudy_data_complete[study_data_complete[\"Mouse ID\"] == \"g989\"]\n", "_____no_output_____" ], [ "# Create a clean DataFrame by dropping the duplicate mouse by its ID.\nclean = study_data_complete[study_data_complete[\"Mouse ID\"] != \"g989\"]\nclean.head()", "_____no_output_____" ], [ "# Checking the number of mice in the clean DataFrame.\nclean['Mouse ID'].nunique()\n", "_____no_output_____" ] ], [ [ "## Summary Statistics", "_____no_output_____" ] ], [ [ "clean", "_____no_output_____" ], [ "# Generate a summary statistics table of mean, median, variance, standard deviation, and SEM of the tumor volume \n# for each regimen\n\ngrp = clean.groupby('Drug Regimen')['Tumor Volume (mm3)']\npd.DataFrame({'mean':grp.mean(),'median':grp.median(),'var':grp.var(),'std':grp.std(),'sem':grp.sem()})", "_____no_output_____" ], [ "# Generate a summary statistics table of mean, median, variance, standard deviation, and SEM of the tumor volume for each regimen\n# Using the aggregation method, produce the same summary statistics in a single line\n\ngrp.agg(['mean','median','var','std','sem'])", "_____no_output_____" ] ], [ [ "## Bar and Pie Charts", "_____no_output_____" ] ], [ [ "# Generate a bar plot showing the total number of measurements taken on each drug regimen using pandas.\n# plot the mouse counts for each drug using pandas\n\nplt.figure(figsize=[15,6])\nmeasurements = clean.groupby('Drug Regimen').Sex.count()\nmeasurements.plot(kind='bar',rot=45,title='Total Measurements per Drug')\nplt.ylabel('Measurements')\nplt.show()", "_____no_output_____" ], [ "measurements", "_____no_output_____" ], [ "measurements.values", "_____no_output_____" ], [ "# Generate a bar plot showing the total number of measurements taken on each drug regimen using pyplot.\n# plot the bar graph of mice count per drug regimen\n\nplt.figure(figsize=[15,6])\nplt.bar(measurements.index,measurements.values)\nplt.title('Total Measurements per Drug')\nplt.ylabel('Measurements')\nplt.xlabel('Drug regimen')\nplt.show()", "_____no_output_____" ], [ "pd.DataFrame.plot()", "_____no_output_____" ], [ "clean.Sex.value_counts().index", "_____no_output_____" ], [ "# Generate a pie plot showing the distribution of female versus male mice using pandas\nclean.Sex.value_counts().plot.pie(autopct='%1.1f%%', explode=[.1,0],shadow=True)\nplt.show()", "_____no_output_____" ], [ "# Generate a pie plot showing the distribution of female versus male mice using pyplot\nplt.pie(clean.Sex.value_counts(), autopct='%1.1f%%',\n labels=clean.Sex.value_counts().index,explode=[.1,0],shadow=True)\nplt.show()", "_____no_output_____" ] ], [ [ "## Quartiles, Outliers and Boxplots", "_____no_output_____" ] ], [ [ "# Reset index so drug regimen column persists after inner merge\n# Start by getting the last (greatest) timepoint for each mouse\ntimemax = clean.groupby('Mouse ID').max().Timepoint.reset_index()\n\n# Merge this group df with the original dataframe to get the tumor volume at the last timepoint\ntumormax = timemax.merge(clean,on=['Mouse ID','Timepoint'])\n# show all rows of data\ntumormax", "_____no_output_____" ], [ "# get mouse count per drug\ntumormax.groupby('Drug Regimen').Timepoint.count()", "_____no_output_____" ], [ "# Calculate the final tumor volume of each mouse across four of the treatment regimens: \n# Capomulin, Ramicane, Infubinol, and Ceftamin\n\n# Put treatments into a list for for loop (and later for plot labels)\ndrugs = ['Capomulin', 'Ramicane', 'Infubinol', 'Ceftamin']\n\n# Create empty list to fill with tumor vol data (for plotting)\ntumor_list = []\n\n# set drug regimen as index and drop associated regimens while only keeping Capomulin, Ramicane, Infubinol, and Ceftamin\nfor drug in drugs:\n # add subset \n # tumor volumes for each Drug Regimen\n # Locate the rows which contain mice on each drug and get the tumor volumes\n tumor_data = tumormax[tumormax['Drug Regimen'] == drug]['Tumor Volume (mm3)']\n # Calculate the IQR and quantitatively determine if there are any potential outliers. \n iqr = tumor_data.quantile(.75) - tumor_data.quantile(.25)\n # Determine outliers using upper and lower bounds\n lower_bound = tumor_data.quantile(.25) - (1.5*iqr)\n upper_bound = tumor_data.quantile(.75) + (1.5*iqr)\n tumor_list.append(tumor_data)\n \n # isolated view of just capomulin for later use\n print(f'{drug} potential outliers: {tumor_data[(tumor_data<lower_bound)|(tumor_data>upper_bound)]}')", "Capomulin potential outliers: Series([], Name: Tumor Volume (mm3), dtype: float64)\nRamicane potential outliers: Series([], Name: Tumor Volume (mm3), dtype: float64)\nInfubinol potential outliers: 31 36.321346\nName: Tumor Volume (mm3), dtype: float64\nCeftamin potential outliers: Series([], Name: Tumor Volume (mm3), dtype: float64)\n" ], [ "# Generate a box plot of the final tumor volume of each mouse across four regimens of interest\nplt.figure(figsize=[10,5])\n#set drugs to be analyzed, colors for the plots, and markers\nplt.boxplot(tumor_list,labels=drugs, flierprops={'markerfacecolor':'red','markersize':30})\nplt.ylabel('Final Tumor Valume (mm3)')\nplt.xticks(fontsize=18)\nplt.show()", "_____no_output_____" ] ], [ [ "## Line and Scatter Plots", "_____no_output_____" ] ], [ [ "# Generate a line plot of tumor volume vs. time point for a mouse treated with Capomulin\n#change index to mouse ID \n#remove other mouse IDs so only s185 shows\n#set the x-axis equal to the Timepoint and y-axis to Tumor Volume\n\nplt.figure(figsize=[15,6])\nclean[(clean['Drug Regimen']=='Capomulin')&(clean['Mouse ID']=='s185')]\\\n.set_index('Timepoint')['Tumor Volume (mm3)'].plot()\nplt.ylabel('Tumor Volume (mm3)')\nplt.title('Tumor Volume vs. Timepoint for Mouse s185')\nplt.grid()\nplt.show()", "_____no_output_____" ], [ "# Generate a scatter plot of average tumor volume vs. mouse weight for the Capomulin regimen\n# group by mouse ID to find average tumor volume\n\ntumor_weight = clean[clean['Drug Regimen']=='Capomulin'].groupby('Mouse ID').mean()\\\n .set_index('Weight (g)')['Tumor Volume (mm3)']\n\n# establish x-axis value for the weight of the mice\n# produce scatter plot of the data\n\nplt.figure(figsize=[15,6])\nplt.scatter(tumor_weight.index,tumor_weight.values)\nplt.xlabel('Weight (g) Average')\nplt.ylabel('Tumor Volume (mm3) Average')\nplt.title('Capomulin Treatment Weight vs Tumor Volume Average')\nplt.show()", "_____no_output_____" ] ], [ [ "## Correlation and Regression", "_____no_output_____" ] ], [ [ "tumor_weight.head()", "_____no_output_____" ], [ "# Calculate the correlation coefficient and linear regression model \n# for mouse weight and average tumor volume for the Capomulin regimen\n#establish x and y values and find St. Pearson Correlation Coefficient for Mouse Weight and Tumor Volume Avg\nlinear_corr = stats.pearsonr(tumor_weight.index,tumor_weight.values)\n\n# establish linear regression values\nmodel = linregress(tumor_weight.index,tumor_weight.values)\n\n# linear regression line \ny_values=tumor_weight.index*model[0]+model[1]\n# scatter plot of the data\nplt.figure(figsize=[15,6])\nplt.plot(tumor_weight.index,y_values,color='red')\nplt.xlabel('Weight (g) Average')\nplt.ylabel('Tumor Volume (mm3) Average')\nplt.title('Capomulin Treatment Weight vs Tumor Volume Average')\nplt.scatter(tumor_weight.index,tumor_weight.values)\nplt.show()\n#print St. Pearson Correlation Coefficient\nprint(f'The correlation between mouse weight and average tumor volume is {linear_corr[0]:.2f}')", "_____no_output_____" ] ] ]

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

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

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "!pip install tkseem", "_____no_output_____" ] ], [ [ "## Question Answering", "_____no_output_____" ], [ "### Download and Prepare Data", "_____no_output_____" ] ], [ [ "!wget https://dl.fbaipublicfiles.com/MLQA/MLQA_V1.zip\n!unzip MLQA_V1.zip", "_____no_output_____" ] ], [ [ "### Prepare Data", "_____no_output_____" ] ], [ [ "import json \n\ndef read_data(file_path, max_context_size = 100):\n # Read dataset\n with open(file_path) as f:\n data = json.load(f)\n\n contexts = []\n questions = []\n answers = []\n labels = []\n \n for i in range(len(data['data'])):\n\n paragraph_object = data['data'][i][\"paragraphs\"]\n \n for j in range(len(paragraph_object)):\n\n context_object = paragraph_object[j]\n context_text = context_object['context']\n\n if len(context_text.split()) > max_context_size:\n continue\n for k in range(len(context_object['qas'])):\n\n question_object = context_object['qas'][k]\n question_text = question_object['question']\n \n answer_object = question_object['answers'][0]\n answer_text = answer_object['text']\n answer_start = answer_object['answer_start']\n answer_end = answer_start + len(answer_text)\n\n answer_start = len(context_text[:answer_start].split())\n answer_end = answer_start + len(answer_text.split())\n if answer_end >= max_context_size:\n answer_end = max_context_size -1\n labels.append([answer_start, answer_end])\n questions.append(question_text)\n contexts.append(context_text)\n answers.append(answer_text)\n \n with open('train_contexts.txt', 'w') as f:\n f.write(('\\n').join(contexts))\n with open('train_questions.txt', 'w') as f:\n f.write(('\\n').join(questions))\n return {'qas':questions, 'ctx':contexts, 'ans':answers, 'lbl':labels}", "_____no_output_____" ], [ "train_data = read_data('MLQA_V1/test/test-context-ar-question-ar.json')", "_____no_output_____" ], [ "for i in range(10):\n print(train_data['qas'][i])\n print(train_data['ctx'][i])\n print(train_data['ans'][i])\n print(\"==============\")", "ما الذي جعل شريط الاختبار للطائرة؟\nبحيرة جرووم كانت تستخدم للقصف المدفعي والتدريب علي المدفعية خلال الحرب العالمية الثانية، ولكن تم التخلي عنها بعد ذلك حتى نيسان / أبريل 1955، عندما تم اختياره من قبل فريق لوكهيد اسكنك كموقع مثالي لاختبار لوكهيد يو-2 - 2 طائرة التجسس. قاع البحيرة قدم الشريط المثالية التي يمكن عمل اختبارات الطائرات المزعجة، ارتفاع سلسلة جبال وادي الإيمجرانت ومحيط NTS يحمي موقع الاختبار من أعين المتطفلين والتدخل الخارجي.\nقاع البحيرة\n==============\nمن كان يرافق طائرة يو -2 عند التسليم؟\nشيدت لوكهيد قاعدة مؤقتة في الموقع، ثم عرفت باسم الموقع الثاني أو \"المزرعة\"، التي تتألف من أكثر بقليل من بضعة مخابئ، وحلقات عمل ومنازل متنقلة لفريقها الصغير. في ثلاثة أشهر فقط شيد مدرج طوله 5000 ودخل الخدمة بحلول تموز / يوليو 1955. حصلت المزرعة على تسليم أول يو 2 في 24 يوليو، 1955 من بوربانك على سي 124 جلوب ماستر الثاني طائرة شحن، يرافقه فنيي وكهيد على دي سي 3. انطلق أول يو - 2 من الجرووم في 4 أغسطس، 1955. بدأت عمليات تحليق أسطول يو 2 تحت سيطرة وكالة المخابرات المركزية الأمريكية في الأجواء السوفياتية بحلول منتصف عام 1956.\nفنيي وكهيد\n==============\n ما هو نوع العمل الذي يواجهه الطيارون العسكريون إذا انتقلوا إلى \\n مناطق محظورة؟ \nعلى عكس الكثير من حدود نيليس، والمنطقة المحيطة بها في البحيرة بشكل دائم خارج الحدود سواء على المدنيين وطبيعية حركة الطيران العسكري. محطات الرادار لحماية المنطقة، والأفراد غير مصرح بها سرعان ما تطرد. حتى طيارين التدريب العسكري في خطر NAFR إجراءات التأديبية إذا تواجدوا بطريق الخطأ في \"المربع\"الحظور للجرووم والأجواء المحيطة بها.\nإجراءات التأديبية\n==============\nمتى تم نشر مقال مجلة الطيران؟\nفي كانون الثاني 2006، نشر مؤرخ الفضاء دواين أ يوم مقال نشر في المجلة الإلكترونية الطيران والفضاء استعراض بعنوان \"رواد الفضاء والمنطقة 51 : حادث سكايلاب\". المقال كان مبنيا على مذكرة مكتوبة في عام 1974 إلى مديروكالة المخابرات المركزية يام كولبي من قبل عملاء مجهولين لوكالة الاستخبارات المركزية. وذكرت المذكرة أن رواد الفضاء على متن سكايلاب 4، وذلك كجزء من برنامج أوسع نطاقا، عن غير قصد بالتقاط صور لموقع الذي قالت المذكرة :\nكانون الثاني 2006\n==============\nما هو الموقع الذي أصبح مركزاً للأطباق الطائرة ونظريات المؤامرة؟\nلطبيعتها السرية وفيما لا شك فيه بحوث تصنيف الطائرات، إلى جانب تقارير عن الظواهر غير العادية، قد أدت الي ان تصبح منطقة 51 مركزا للاطباق الطائرة الحديثة ونظريات المؤامرة. بعض الأنشطة المذكورة في مثل هذه النظريات في منطقة 51 تشمل ما يلي :\nمنطقة 51\n==============\nما كان محور مؤامرة الجسم الغريب الحديثة؟\\n\nلطبيعتها السرية وفيما لا شك فيه بحوث تصنيف الطائرات، إلى جانب تقارير عن الظواهر غير العادية، قد أدت الي ان تصبح منطقة 51 مركزا للاطباق الطائرة الحديثة ونظريات المؤامرة. بعض الأنشطة المذكورة في مثل هذه النظريات في منطقة 51 تشمل ما يلي :\nمنطقة 51\n==============\nمالذي يُظن بأنه قد تم بنائه في روزويل؟\nالتخزين، والفحص، والهندسة العكسية للمركبة الفضائية الغريبة المحطمة (بما في ذلك مواد يفترض ان تعافى في روزويل)، ودراسة شاغليها (حية أو ميتة)، وصناعة الطائرات على أساس التكنولوجيا الغريبة.\nصناعة الطائرات على أساس التكنولوجيا الغريبة\n==============\nمتى يقوم Qos بالتفاوض على كيفية عمل الشبكة؟\nويمكن أن تتوافق الشبكة أو البروتوكول الذي يدعم جودة الخدمات على عقد المرور مع تطبيق البرمجيات والقدرة الاحتياطية في عقد الشبكة، على سبيل المثال خلال مرحلة إقامة الدورات. وهي يمكن أن تحقق رصدا لمستوى الأداء خلال الدورة، على سبيل المثال معدل البيانات والتأخير، والتحكم ديناميكيا عن طريق جدولة الأولويات في عقد الشبكة. وقد تفرج عن القدرة الاحتياطية خلال مرحلة الهدم.\nمرحلة إقامة الدورات\n==============\nما هو أحد الشروط للتجارة الشبكية المتنوعة؟\nجودة الخدمة قد تكون مطلوبة لأنواع معينة من حركة مرور الشبكة، على سبيل المثال :\nجودة الخدمة\n==============\nكم عدد قوائم الانتظار الموجودة على أجهزة توجيه المختلفة؟\\n\nالموجهات لدعم DiffServ استخدام قوائم متعددة للحزم في انتظار انتقال من عرض النطاق الترددي مقيدة (على سبيل المثال، منطقة واسعة) واجهات. راوتر الباعة يوفر قدرات مختلفة لتكوين هذا السلوك، لتشمل عددا من قوائم معتمدة، والأولويات النسبية لقوائم الانتظار، وعرض النطاق الترددي المخصصة لكل قائمة انتظار.\nمتعددة\n==============\n" ] ], [ [ "### Imports", "_____no_output_____" ] ], [ [ "import re\nimport nltk\nimport time\nimport numpy as np\nimport tkseem as tk\nimport tensorflow as tf\nimport matplotlib.ticker as ticker\nimport matplotlib.pyplot as plt", "_____no_output_____" ] ], [ [ "### Tokenization", "_____no_output_____" ] ], [ [ "qa_tokenizer = tk.WordTokenizer()\nqa_tokenizer.train('train_questions.txt')\nprint('Vocab size ', qa_tokenizer.vocab_size)\n\ncx_tokenizer = tk.WordTokenizer()\ncx_tokenizer.train('train_contexts.txt')\nprint('Vocab size ', cx_tokenizer.vocab_size)\n\ntrain_inp_data = qa_tokenizer.encode_sentences(train_data['qas'])\ntrain_tar_data = cx_tokenizer.encode_sentences(train_data['ctx'])\ntrain_tar_lbls = train_data['lbl']\ntrain_inp_data.shape, train_tar_data.shape", "Training WordTokenizer ...\nVocab size 8883\nTraining WordTokenizer ...\nVocab size 10000\n" ] ], [ [ "### Create Dataset", "_____no_output_____" ] ], [ [ "BATCH_SIZE = 64\nBUFFER_SIZE = len(train_inp_data)\n\ndataset = tf.data.Dataset.from_tensor_slices((train_inp_data, train_tar_data, train_tar_lbls)).shuffle(BUFFER_SIZE)\ndataset = dataset.batch(BATCH_SIZE, drop_remainder=True)", "_____no_output_____" ] ], [ [ "### Create Encoder and Decoder", "_____no_output_____" ] ], [ [ "class Encoder(tf.keras.Model):\n def __init__(self, vocab_size, embedding_dim, enc_units, batch_sz):\n super(Encoder, self).__init__()\n self.batch_sz = batch_sz\n self.enc_units = enc_units\n self.embedding = tf.keras.layers.Embedding(vocab_size, embedding_dim)\n self.gru = tf.keras.layers.GRU(self.enc_units,\n recurrent_initializer='glorot_uniform')\n\n def call(self, x, hidden):\n x = self.embedding(x)\n output = self.gru(x, initial_state = hidden)\n return output\n\n def initialize_hidden_state(self):\n return tf.zeros((self.batch_sz, self.enc_units))\n \nclass Decoder(tf.keras.Model):\n def __init__(self, vocab_size, embedding_dim, dec_units, output_sz):\n super(Decoder, self).__init__()\n self.dec_units = dec_units\n self.embedding_dim = embedding_dim\n self.output_sz = output_sz\n self.embedding = tf.keras.layers.Embedding(vocab_size, embedding_dim)\n self.gru = tf.keras.layers.GRU(self.dec_units,\n return_sequences=False,\n recurrent_initializer='glorot_uniform')\n self.fc11 = tf.keras.layers.Dense(embedding_dim)\n self.fc12 = tf.keras.layers.Dense(output_sz)\n\n self.fc21 = tf.keras.layers.Dense(embedding_dim)\n self.fc22 = tf.keras.layers.Dense(output_sz)\n \n def call(self, x, hidden):\n x = self.embedding(x)\n x = self.gru(x, initial_state = hidden)\n x1 = self.fc11(x)\n x2 = self.fc21(x)\n\n x1 = self.fc12(x1)\n x2 = self.fc22(x2)\n return [x1, x2]\n\ndef loss_fn(true, pred):\n cross_entropy = tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True)\n return (cross_entropy(true[:,0:1], pred[0]) + cross_entropy(true[:,1:2], pred[1]))/2", "_____no_output_____" ] ], [ [ "### Training", "_____no_output_____" ] ], [ [ "units = 1024\nembedding_dim = 256\nmax_length_inp = train_inp_data.shape[1]\nmax_length_tar = train_tar_data.shape[1]\nvocab_tar_size = cx_tokenizer.vocab_size\nvocab_inp_size = qa_tokenizer.vocab_size\nsteps_per_epoch = len(train_inp_data) // BATCH_SIZE\ndecoder = Decoder(vocab_tar_size, embedding_dim, units, max_length_tar)\nencoder = Encoder(vocab_inp_size, embedding_dim, units, BATCH_SIZE)\n\noptim = tf.optimizers.Adam()", "_____no_output_____" ], [ "epochs = 25\nfor epoch in range(epochs):\n enc_hidden = encoder.initialize_hidden_state()\n epoch_loss = 0\n \n for idx, (inp, tar, true) in enumerate(dataset):\n with tf.GradientTape() as tape:\n hidden = encoder(inp, enc_hidden)\n pred = decoder(tar, hidden)\n loss = loss_fn(true, pred)\n variables = decoder.trainable_variables + encoder.trainable_variables\n gradients = tape.gradient(loss, variables)\n optim.apply_gradients(zip(gradients, variables))\n epoch_loss += loss.numpy()\n print(f\"Epoch {epoch} loss: {epoch_loss/steps_per_epoch:.3f}\")", "Epoch 0 loss: 4.386\nEpoch 1 loss: 4.264\nEpoch 2 loss: 4.238\nEpoch 3 loss: 4.105\nEpoch 4 loss: 3.932\nEpoch 5 loss: 3.758\nEpoch 6 loss: 3.643\nEpoch 7 loss: 3.548\nEpoch 8 loss: 3.456\nEpoch 9 loss: 3.382\nEpoch 10 loss: 3.285\nEpoch 11 loss: 3.215\nEpoch 12 loss: 3.141\nEpoch 13 loss: 3.047\nEpoch 14 loss: 2.916\nEpoch 15 loss: 2.831\nEpoch 16 loss: 2.748\nEpoch 17 loss: 2.614\nEpoch 18 loss: 2.462\nEpoch 19 loss: 2.306\nEpoch 20 loss: 2.126\nEpoch 21 loss: 1.944\nEpoch 22 loss: 1.770\nEpoch 23 loss: 1.637\nEpoch 24 loss: 1.414\n" ] ], [ [ "### evaluation", "_____no_output_____" ] ], [ [ "def answer(question_txt, context_txt, answer_txt_tru):\n question = qa_tokenizer.encode_sentences([question_txt], out_length = max_length_inp)\n context = cx_tokenizer.encode_sentences([context_txt], out_length = max_length_tar)\n question = tf.convert_to_tensor(question)\n context = tf.convert_to_tensor(context)\n result = ''\n\n hidden = [tf.zeros((1, units))]\n enc_hidden = encoder(question, hidden)\n pred = decoder(context, enc_hidden)\n\n start = tf.argmax(pred[0], axis = -1).numpy()[0]\n end = tf.argmax(pred[1], axis = -1).numpy()[0]\n \n if start >= len(context_txt.split()):\n start = len(context_txt.split()) - 1\n if end >= len(context_txt.split()):\n end = len(context_txt.split()) - 1\n \n # if one word prediction\n if end == start:\n end += 1\n answer_txt = (' ').join(context_txt.split()[start:end])\n \n print(\"Question : \", question_txt)\n print(\"Context : \",context_txt)\n print(\"Pred Answer : \",answer_txt)\n print(\"True Answer : \", answer_txt_tru)\n print(\"======================\")", "_____no_output_____" ], [ "answer(\"في أي عام توفي وليام ؟\", \"توفي وليام في عام 1990\", \"1990\")\nanswer(\"ماهي عاصمة البحرين ؟\", \"عاصمة البحرين هي المنامة\", \"المنامة\")\nanswer(\"في أي دولة ولد جون ؟\", \"ولد في فرنسا عام 1988\", \"فرنسا\")\nanswer(\"أين تركت الهاتف ؟\", \"تركت الهاتف فوق الطاولة\", \"فوق الطاولة\")", "Question : في أي عام توفي وليام ؟\nContext : توفي وليام في عام 1990\nPred Answer : 1990\nTrue Answer : 1990\n======================\nQuestion : ماهي عاصمة البحرين ؟\nContext : عاصمة البحرين هي المنامة\nPred Answer : المنامة\nTrue Answer : المنامة\n======================\nQuestion : في أي دولة ولد جون ؟\nContext : ولد في فرنسا عام 1988\nPred Answer : 1988\nTrue Answer : فرنسا\n======================\nQuestion : أين تركت الهاتف ؟\nContext : تركت الهاتف فوق الطاولة\nPred Answer : الطاولة\nTrue Answer : فوق الطاولة\n======================\n" ] ] ]

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

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

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import matplotlib.pyplot as plt\nfrom IPython.display import set_matplotlib_formats\nset_matplotlib_formats('retina', 'png')\n\nimport numpy as np\n\nimport seaborn as sns\nsns.set_style('white')\n\nimport mpld3\n\nimport mygene\nfrom geneinfo import geneinfo, geneplot, connect_mygene\n\nmg = mygene.MyGeneInfo()\nconnect_mygene(mg)", "/Users/kmt/anaconda3/envs/simons/lib/python3.6/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" ] ], [ [ "## Single gene name", "_____no_output_____" ] ], [ [ "geneinfo('USP4')", "_____no_output_____" ] ], [ [ "## List of names", "_____no_output_____" ] ], [ [ "geneinfo(['LARS2', 'XCR1'])", "_____no_output_____" ] ], [ [ "## Get all protein coding genes in a (hg38) region", "_____no_output_____" ] ], [ [ "for gene in mg.query('q=chr2:49500000-50000000 AND type_of_gene:protein-coding', species='human', fetch_all=True):\n geneinfo(gene['symbol'])", "Fetching 4 gene(s) . . .\n" ] ], [ [ "## Plot data over gene annotation", "_____no_output_____" ] ], [ [ "chrom, start, end = 'chr3', 49500000, 50600000\nax = geneplot(chrom, start, end, figsize=(10, 5))\n\nax.plot(np.linspace(start, end, 1000), np.random.random(1000), 'o') ;\n\nmpld3.display()", "Fetching 61 gene(s) . . .\n" ], [ "geneinfo(['HYAL3', 'IFRD2'])", "_____no_output_____" ] ] ]

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

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

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Convolutional Neural Networks\n\nIn this notebook we will implement a convolutional neural network. Rather than doing everything from scratch we will make use of [TensorFlow 2](https://www.tensorflow.org/) and the [Keras](https://keras.io) high level interface.", "_____no_output_____" ], [ "## Installing TensorFlow and Keras\n\nTensorFlow and Keras are not included with the base Anaconda install, but can be easily installed by running the following commands on the Anaconda Command Prompt/terminal window:\n```\nconda install notebook jupyterlab nb_conda_kernels\nconda create -n tf tensorflow ipykernel mkl\n```\nOnce this has been done, you should be able to select the `Python [conda env:tf]` kernel from the Kernel->Change Kernel menu item at the top of this notebook. Then, we import TensorFlow package:", "_____no_output_____" ] ], [ [ "import tensorflow as tf", "_____no_output_____" ] ], [ [ "## Creating a simple network with TensorFlow\n\nWe will start by creating a very simple fully connected feedforward network using TensorFlow/Keras. The network will mimic the one we implemented previously, but TensorFlow/Keras will take care of most of the details for us.\n\n### MNIST Dataset\n\nFirst, let us load the MNIST digits dataset that we will be using to train our network. This is available directly within Keras:", "_____no_output_____" ] ], [ [ "(x_train, y_train),(x_test, y_test) = tf.keras.datasets.mnist.load_data()", "_____no_output_____" ] ], [ [ "The data comes as a set of integers in the range [0,255] representing the shade of gray of a given pixel. Let's first rescale them to be in the range [0,1]:", "_____no_output_____" ] ], [ [ "x_train, x_test = x_train / 255.0, x_test / 255.0", "_____no_output_____" ] ], [ [ "Now we can build a neural network model using Keras. This uses a very simple high-level modular structure where we only have the specify the layers in our model and the properties of each layer. The layers we will have are as follows:\n1. Input layer: This will be a 28x28 matrix of numbers.\n2. `Flatten` layer: Convert our 28x28 pixel image into an array of size 784.\n3. `Dense` layer: a fully-connected layer of the type we have been using up to now. We will use 30 neurons and the sigmoid activation function.\n4. `Dense` layer: fully-connected output layer. ", "_____no_output_____" ] ], [ [ "model = tf.keras.models.Sequential([\n tf.keras.layers.Flatten(input_shape=(28, 28)),\n tf.keras.layers.Dense(30, activation='sigmoid'),\n tf.keras.layers.Dense(10, activation='softsign')\n])", "_____no_output_____" ], [ "model.compile(optimizer='adam',\n loss='mean_squared_logarithmic_error',\n metrics=['accuracy'])", "_____no_output_____" ], [ "model.fit(x_train, y_train, epochs=5)", "Epoch 1/5\n1875/1875 [==============================] - 1s 776us/step - loss: 1.7309 - accuracy: 0.0955\nEpoch 2/5\n1875/1875 [==============================] - 2s 840us/step - loss: 1.6526 - accuracy: 0.0987\nEpoch 3/5\n1875/1875 [==============================] - 2s 892us/step - loss: 1.6396 - accuracy: 0.09870s - loss: 1.6406 - accuracy: 0.09\nEpoch 4/5\n1875/1875 [==============================] - 2s 1ms/step - loss: 1.6327 - accuracy: 0.0987: 0s - los\nEpoch 5/5\n1875/1875 [==============================] - 2s 1ms/step - loss: 1.6282 - accuracy: 0.0987\n" ], [ "model.evaluate(x_test, y_test)", "313/313 [==============================] - 0s 719us/step - loss: 0.1253 - accuracy: 0.9629\n" ] ], [ [ "#### Exercises\nExperiment with this network:\n1. Change the number of neurons in the hidden layer.\n2. Add more hidden layers.\n3. Change the activation function in the hidden layer to `relu`.\n4. Change the activation in the output layer to `softmax`.\nHow does the performance of your network change with these modifications?\n\n#### Task\nImplement the neural network in \"[Gradient-based learning applied to document recognition](http://yann.lecun.com/exdb/publis/pdf/lecun-98.pdf)\", by Yann LeCun, Léon Bottou, Yoshua Bengio, and Patrick Haffner.", "_____no_output_____" ] ], [ [ "model = tf.keras.models.Sequential([\n tf.keras.layers.Conv2D(6,5, activation='sigmoid',input_shape=(28, 28,1)),\n tf.keras.layers.MaxPooling2D(pool_size=(2,2)),\n tf.keras.layers.Conv2D(16,5, activation='sigmoid'),\n tf.keras.layers.MaxPooling2D(pool_size=(2,2)),\n tf.keras.layers.Flatten(),\n tf.keras.layers.Dense(84, activation='sigmoid'),\n tf.keras.layers.Dense(10, activation='softmax')\n])", "_____no_output_____" ], [ "model.compile(optimizer='adam',\n loss='sparse_categorical_crossentropy',\n metrics=['accuracy'])", "_____no_output_____" ], [ "model.fit(x_train, y_train, epochs=5)", "Epoch 1/5\n" ] ] ]

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

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

[ "Unlicense" ]

[ "Unlicense" ]

[ "Unlicense" ]

[ [ [ "## Introduction", "_____no_output_____" ], [ "This notebook was intended to show the nuclear isotope abundance dependent on baryon density. I abandoned it later though. If you want to pick it up, go ahead. No guarantees all rights reserved and you are responsible for your interpretation of all this, etc.", "_____no_output_____" ], [ "## Imports", "_____no_output_____" ] ], [ [ "import numpy as np\nimport matplotlib.pyplot as plt\n%matplotlib inline\nplt.rc('font', size=18)\nplt.rcParams['figure.figsize'] = (10.0, 7.0)", "_____no_output_____" ] ], [ [ "## Read data", "_____no_output_____" ] ], [ [ "def read_datathief(fn):\n data = np.loadtxt(fn, converters={0: lambda x: x[:-1]})\n return data[:, 0], data[:, 1]", "_____no_output_____" ], [ "ab = {}\nelems = ['he4', 'd2', 'he3', 'li3']\nab['x_he4'], ab['he4'] = read_datathief('data/abundances/He4.txt')\nab['x_d2'], ab['d2'] = read_datathief('data/abundances/D2.txt')\nab['x_he3'], ab['he3'] = read_datathief('data/abundances/He3.txt')\nab['x_li3'], ab['li3'] = read_datathief('data/abundances/Li.txt')", "_____no_output_____" ], [ "ab['li3_c'] = 1.58e-10\nab['li3_c_d'] = 0.3e-10\nab['d2_c'] = 2.53e-5\nab['d2_c_d'] = 0.04e-5\nab['he3_c'] = 1.1e-5\nab['he3_c_d'] = 0.2e-5\nab['he4_c'] = 0.2449\nab['he4_c_d'] = 0.004\n", "_____no_output_____" ] ], [ [ "## Plots", "_____no_output_____" ], [ "### All on one", "_____no_output_____" ] ], [ [ "for el in elems:\n plt.plot(ab['x_' + el], ab[el])\nplt.xscale('log')\nplt.yscale('log')\nplt.xlabel('Baryon fraction')\nplt.ylabel('Abundance')\n# plt.ylim(1e-10, 1)", "_____no_output_____" ] ], [ [ "### Fancy plot", "_____no_output_____" ] ], [ [ "f, (ax1, ax2, ax3) = plt.subplots(3, 1, sharex=True)\naxs = [ax1, ax2, ax3]\nc = {el: 'C%d' % i for i, el in enumerate(elems) }\nplanck_ab = 0.02230\nfor el in elems:\n for ax in axs:\n ax.plot(ab['x_' + el], ab[el], color=c[el], lw=3)\n ax.axhline(ab[el+'_c'], color=c[el])\n ax.axvline(planck_ab)\nax1.set_ylim(0.22, 0.26)\nax2.set_ylim(1e-6, 1e-3)\nax3.set_ylim(1e-10, 1e-8)\nfor ax in axs:\n ax.set_yscale('log')\n ax.set_xscale('log')\n\nfor ax in (ax1, ax2):\n ax.spines['bottom'].set_visible(False)\n\nax = ax1\nax.xaxis.tick_bottom()\nax.xaxis.tick_top()\nax.tick_params(labeltop='off')\n\nplt.sca(ax2)\nplt.tick_params(\n axis='x', # changes apply to the x-axis\n which='both', # both major and minor ticks are affected\n bottom='off', # ticks along the bottom edge are off\n top='off', # ticks along the top edge are off\n labelbottom='off')\n\nfor ax in (ax2, ax3):\n ax.spines['top'].set_visible(False)\n# ax.tick_params(labeltop='off') # don't put tick labels at the top\n\n\n\nd = .01 # how big to make the diagonal lines in axes coordinates\n# arguments to pass to plot, just so we don't keep repeating them\nax = ax1\nkwargs = dict(transform=ax.transAxes, color='k', clip_on=False)\nax.plot((-d, +d), (-d, +d), **kwargs) # top-left diagonal\nax.plot((1 - d, 1 + d), (-d, +d), **kwargs) # top-right diagonal\n\nax = ax2\nkwargs = dict(transform=ax.transAxes, color='k', clip_on=False)\nax.plot((-d, +d), (-d, +d), **kwargs) # top-left diagonal\nax.plot((1 - d, 1 + d), (-d, +d), **kwargs) # top-right diagonal\n\nkwargs.update(transform=ax.transAxes) # switch to the bottom axes\nax.plot((-d, +d), (1 - d, 1 + d), **kwargs) # bottom-left diagonal\nax.plot((1 - d, 1 + d), (1 - d, 1 + d), **kwargs) # bottom-right diagonal\n\nax = ax3\nkwargs.update(transform=ax.transAxes) # switch to the bottom axes\nax.plot((-d, +d), (1 - d, 1 + d), **kwargs) # bottom-left diagonal\nax.plot((1 - d, 1 + d), (1 - d, 1 + d), **kwargs) # bottom-right diagonal\n\nplt.xlim(4e-3, 3e-2)", "_____no_output_____" ] ] ]

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

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

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ [ [ "<img src=\"../../img/logo_white_bkg_small.png\" align=\"left\" /> \n# Supervised Learning\nThis worksheet covers concepts covered in the second part of day 2 - Feature Engineering. It should take no more than 40-60 minutes to complete. Please raise your hand if you get stuck. \n\n## Import the Libraries\nFor this exercise, we will be using:\n* Pandas (http://pandas.pydata.org/pandas-docs/stable/)\n* Numpy (https://docs.scipy.org/doc/numpy/reference/)\n* Matplotlib (http://matplotlib.org/api/pyplot_api.html)\n* Scikit-learn (http://scikit-learn.org/stable/documentation.html)\n* YellowBrick (http://www.scikit-yb.org/en/latest/)\n* Seaborn (https://seaborn.pydata.org)\n* Lime (https://github.com/marcotcr/lime)", "_____no_output_____" ] ], [ [ "# Load Libraries - Make sure to run this cell!\nimport pandas as pd\nimport numpy as np\nimport re\nfrom collections import Counter\nfrom sklearn import feature_extraction, tree, model_selection, metrics\nfrom yellowbrick.classifier import ClassificationReport\nfrom yellowbrick.classifier import ConfusionMatrix\nimport matplotlib.pyplot as plt\nimport matplotlib\nimport lime\n%matplotlib inline", "_____no_output_____" ] ], [ [ "## Worksheet - DGA Detection using Machine Learning\n\nThis worksheet is a step-by-step guide on how to detect domains that were generated using \"Domain Generation Algorithm\" (DGA). We will walk you through the process of transforming raw domain strings to Machine Learning features and creating a decision tree classifer which you will use to determine whether a given domain is legit or not. Once you have implemented the classifier, the worksheet will walk you through evaluating your model. \n\nOverview 2 main steps:\n\n1. **Feature Engineering** - from raw domain strings to numeric Machine Learning features using DataFrame manipulations\n2. **Machine Learning Classification** - predict whether a domain is legit or not using a Decision Tree Classifier\n\n\n \n\n**DGA - Background**\n\n\"Various families of malware use domain generation\nalgorithms (DGAs) to generate a large number of pseudo-random\ndomain names to connect to a command and control (C2) server.\nIn order to block DGA C2 traffic, security organizations must\nfirst discover the algorithm by reverse engineering malware\nsamples, then generate a list of domains for a given seed. The\ndomains are then either preregistered, sink-holed or published\nin a DNS blacklist. This process is not only tedious, but can\nbe readily circumvented by malware authors. An alternative\napproach to stop malware from using DGAs is to intercept DNS\nqueries on a network and predict whether domains are DGA\ngenerated. Much of the previous work in DGA detection is based\non finding groupings of like domains and using their statistical\nproperties to determine if they are DGA generated. However,\nthese techniques are run over large time windows and cannot be\nused for real-time detection and prevention. In addition, many of\nthese techniques also use contextual information such as passive\nDNS and aggregations of all NXDomains throughout a network.\nSuch requirements are not only costly to integrate, they may not\nbe possible due to real-world constraints of many systems (such\nas endpoint detection). An alternative to these systems is a much\nharder problem: detect DGA generation on a per domain basis\nwith no information except for the domain name. Previous work\nto solve this harder problem exhibits poor performance and many\nof these systems rely heavily on manual creation of features;\na time consuming process that can easily be circumvented by\nmalware authors...\" \n[Citation: Woodbridge et. al 2016: \"Predicting Domain Generation Algorithms with Long Short-Term Memory Networks\"]\n\nA better alternative for real-world deployment would be to use \"featureless deep learning\" - We have a separate notebook where you can see how this can be implemented!\n\n**However, let's learn the basics first!!!**\n", "_____no_output_____" ], [ "## Feature Engineering", "_____no_output_____" ], [ "#### Breakpoint: Load Features and Labels\n\nIf you got stuck in Part 1, please simply load the feature matrix we prepared for you, so you can move on to Part 2 and train a Decision Tree Classifier.", "_____no_output_____" ] ], [ [ "df_final = pd.read_csv('../../Data/dga_features_final_df.csv')\nprint(df_final.isDGA.value_counts())\ndf_final.head()", "_____no_output_____" ], [ "# Load dictionary of common english words from part 1\nfrom six.moves import cPickle as pickle\nwith open('../../Data/d_common_en_words' + '.pickle', 'rb') as f:\n d = pickle.load(f)", "_____no_output_____" ] ], [ [ "## Part 2 - Machine Learning\n\nTo learn simple classification procedures using [sklearn](http://scikit-learn.org/stable/) we have split the work flow into 5 steps.", "_____no_output_____" ], [ "### Step 1: Prepare Feature matrix and ```target``` vector containing the URL labels\n\n- In statistics, the feature matrix is often referred to as ```X```\n- target is a vector containing the labels for each URL (often also called *y* in statistics)\n- In sklearn both the input and target can either be a pandas DataFrame/Series or numpy array/vector respectively (can't be lists!)\n\nTasks:\n- assign 'isDGA' column to a pandas Series named 'target'\n- drop 'isDGA' column from ```dga``` DataFrame and name the resulting pandas DataFrame 'feature_matrix'", "_____no_output_____" ] ], [ [ "#Your code here ...", "_____no_output_____" ] ], [ [ "### Step 2: Simple Cross-Validation\n\nTasks:\n- split your feature matrix X and target vector into train and test subsets using sklearn [model_selection.train_test_split](http://scikit-learn.org/stable/modules/generated/sklearn.model_selection.train_test_split.html)", "_____no_output_____" ] ], [ [ "# Simple Cross-Validation: Split the data set into training and test data\n#Your code here ...", "_____no_output_____" ] ], [ [ "### Step 3: Train the model and make a prediction\n\nFinally, we have prepared and segmented the data. Let's start classifying!! \n\nTasks:\n\n- Use the sklearn [tree.DecisionTreeClassfier()](http://scikit-learn.org/stable/modules/generated/sklearn.tree.DecisionTreeClassifier.html), create a decision tree with standard parameters, and train it using the ```.fit()``` function with ```X_train``` and ```target_train``` data.\n- Next, pull a few random rows from the data and see if your classifier got it correct.\n\nIf you are interested in trying a real unknown domain, you'll have to create a function to generate the features for that domain before you run it through the classifier (see function ```is_dga``` a few cells below). ", "_____no_output_____" ] ], [ [ "# Train the decision tree based on the entropy criterion\n\n#Your code here ...", "_____no_output_____" ], [ "# For simplicity let's just copy the needed function in here again\n\ndef H_entropy (x):\n # Calculate Shannon Entropy\n prob = [ float(x.count(c)) / len(x) for c in dict.fromkeys(list(x)) ] \n H = - sum([ p * np.log2(p) for p in prob ]) \n return H\n\ndef vowel_consonant_ratio (x):\n # Calculate vowel to consonant ratio\n x = x.lower()\n vowels_pattern = re.compile('([aeiou])')\n consonants_pattern = re.compile('([b-df-hj-np-tv-z])')\n vowels = re.findall(vowels_pattern, x)\n consonants = re.findall(consonants_pattern, x)\n try:\n ratio = len(vowels) / len(consonants)\n except: # catch zero devision exception \n ratio = 0 \n return ratio\n\n# ngrams: Implementation according to Schiavoni 2014: \"Phoenix: DGA-based Botnet Tracking and Intelligence\"\n# http://s2lab.isg.rhul.ac.uk/papers/files/dimva2014.pdf\n\ndef ngrams(word, n):\n # Extract all ngrams and return a regular Python list\n # Input word: can be a simple string or a list of strings\n # Input n: Can be one integer or a list of integers \n # if you want to extract multipe ngrams and have them all in one list\n \n l_ngrams = []\n if isinstance(word, list):\n for w in word:\n if isinstance(n, list):\n for curr_n in n:\n ngrams = [w[i:i+curr_n] for i in range(0,len(w)-curr_n+1)]\n l_ngrams.extend(ngrams)\n else:\n ngrams = [w[i:i+n] for i in range(0,len(w)-n+1)]\n l_ngrams.extend(ngrams)\n else:\n if isinstance(n, list):\n for curr_n in n:\n ngrams = [word[i:i+curr_n] for i in range(0,len(word)-curr_n+1)]\n l_ngrams.extend(ngrams)\n else:\n ngrams = [word[i:i+n] for i in range(0,len(word)-n+1)]\n l_ngrams.extend(ngrams)\n# print(l_ngrams)\n return l_ngrams\n\ndef ngram_feature(domain, d, n):\n # Input is your domain string or list of domain strings\n # a dictionary object d that contains the count for most common english words\n # finally you n either as int list or simple int defining the ngram length\n \n # Core magic: Looks up domain ngrams in english dictionary ngrams and sums up the \n # respective english dictionary counts for the respective domain ngram\n # sum is normalized\n \n l_ngrams = ngrams(domain, n)\n# print(l_ngrams)\n count_sum=0\n for ngram in l_ngrams:\n if d[ngram]:\n count_sum+=d[ngram]\n try:\n feature = count_sum/(len(domain)-n+1)\n except:\n feature = 0\n return feature\n \ndef average_ngram_feature(l_ngram_feature):\n # input is a list of calls to ngram_feature(domain, d, n)\n # usually you would use various n values, like 1,2,3...\n return sum(l_ngram_feature)/len(l_ngram_feature)", "_____no_output_____" ], [ "def is_dga(domain, clf, d):\n # Function that takes new domain string, trained model 'clf' as input and\n # dictionary d of most common english words\n # returns prediction\n \n domain_features = np.empty([1,5])\n # order of features is ['length', 'digits', 'entropy', 'vowel-cons', 'ngrams']\n domain_features[0,0] = len(domain)\n pattern = re.compile('([0-9])')\n domain_features[0,1] = len(re.findall(pattern, domain))\n domain_features[0,2] = H_entropy(domain)\n domain_features[0,3] = vowel_consonant_ratio(domain)\n domain_features[0,4] = average_ngram_feature([ngram_feature(domain, d, 1), \n ngram_feature(domain, d, 2), \n ngram_feature(domain, d, 3)])\n pred = clf.predict(domain_features)\n return pred[0]\n\n\nprint('Predictions of domain %s is [0 means legit and 1 dga]: ' %('spardeingeld'), is_dga('spardeingeld', clf, d)) \nprint('Predictions of domain %s is [0 means legit and 1 dga]: ' %('google'), is_dga('google', clf, d)) \nprint('Predictions of domain %s is [0 means legit and 1 dga]: ' %('1vxznov16031kjxneqjk1rtofi6'), is_dga('1vxznov16031kjxneqjk1rtofi6', clf, d)) \nprint('Predictions of domain %s is [0 means legit and 1 dga]: ' %('lthmqglxwmrwex'), is_dga('lthmqglxwmrwex', clf, d)) \n", "_____no_output_____" ] ], [ [ "### Step 4: Assess model accuracy with simple cross-validation\n\nTasks:\n- Make predictions for all your data. Call the ```.predict()``` method on the clf with your training data ```X_train``` and store the results in a variable called ```target_pred```.\n- Use sklearn [metrics.accuracy_score](http://scikit-learn.org/stable/modules/generated/sklearn.metrics.accuracy_score.html) to determine your models accuracy. Detailed Instruction:\n - Use your trained model to predict the labels of your test data ```X_test```. Run ```.predict()``` method on the clf with your test data ```X_test``` and store the results in a variable called ```target_pred```.. \n - Then calculate the accuracy using ```target_test``` (which are the true labels/groundtruth) AND your models predictions on the test portion ```target_pred``` as inputs. The advantage here is to see how your model performs on new data it has not been seen during the training phase. The fair approach here is a simple **cross-validation**!\n \n- Print out the confusion matrix using [metrics.confusion_matrix](http://scikit-learn.org/stable/modules/generated/sklearn.metrics.confusion_matrix.html)\n- Use Yellowbrick to visualize the classification report and confusion matrix. (http://www.scikit-yb.org/en/latest/examples/modelselect.html#common-metrics-for-evaluating-classifiers)", "_____no_output_____" ] ], [ [ "# fair approach: make prediction on test data portion\n#Your code here ...", "_____no_output_____" ], [ "# Classification Report...neat summary\n#Your code here ...", "_____no_output_____" ] ], [ [ "### Step 5: Assess model accuracy with k-fold cross-validation\n\nTasks:\n- Partition the dataset into *k* different subsets\n- Create *k* different models by training on *k-1* subsets and testing on the remaining subsets\n- Measure the performance on each of the models and take the average measure.\n\n*Short-Cut*\nAll of these steps can be easily achieved by simply using sklearn's [model_selection.KFold()](http://scikit-learn.org/stable/modules/generated/sklearn.model_selection.KFold.html) and [model_selection.cross_val_score()](http://scikit-learn.org/stable/modules/generated/sklearn.model_selection.cross_val_score.html) functions.", "_____no_output_____" ] ], [ [ "#Your code here ...", "_____no_output_____" ] ], [ [ "#### (Optional) Visualizing your Tree\nAs an optional step, you can actually visualize your tree. The following code will generate a graph of your decision tree. You will need graphviz (http://www.graphviz.org) and pydotplus (or pydot) installed for this to work.\nThe Griffon VM has this installed already, but if you try this on a Mac, or Linux machine you will need to install graphviz.", "_____no_output_____" ] ], [ [ "# These libraries are used to visualize the decision tree and require that you have GraphViz\n# and pydot or pydotplus installed on your computer.\n\nfrom sklearn.externals.six import StringIO \nfrom IPython.core.display import Image\nimport pydotplus as pydot\n\n\ndot_data = StringIO() \ntree.export_graphviz(clf, out_file=dot_data, \n feature_names=['length', 'digits', 'entropy', 'vowel-cons', 'ngrams'],\n filled=True, rounded=True, \n special_characters=True) \n\ngraph = pydot.graph_from_dot_data(dot_data.getvalue()) \nImage(graph.create_png())\n", "_____no_output_____" ] ], [ [ "## Other Models\nNow that you've built a Decision Tree, let's try out two other classifiers and see how they perform on this data. For this next exercise, create classifiers using:\n\n* Support Vector Machine\n* Random Forest\n* K-Nearest Neighbors (http://scikit-learn.org/stable/modules/generated/sklearn.neighbors.KNeighborsClassifier.html) \n\nOnce you've done that, run the various performance metrics to determine which classifier works best.", "_____no_output_____" ] ], [ [ "from sklearn import svm\nfrom sklearn.ensemble import RandomForestClassifier\nfrom sklearn.neighbors import KNeighborsClassifier", "_____no_output_____" ], [ "#Create the Random Forest Classifier\n", "_____no_output_____" ], [ "#Next, create the SVM classifier\n", "_____no_output_____" ], [ "#Finally the knn\n", "_____no_output_____" ] ], [ [ "## Explain a Prediction\nIn the example below, you can use LIME to explain how a classifier arrived at its prediction. Try running LIME with the various classifiers you've created and various rows to see how it functions. ", "_____no_output_____" ] ], [ [ "import lime.lime_tabular\nexplainer = lime.lime_tabular.LimeTabularExplainer(feature_matrix_train, \n feature_names=['length', 'digits', 'entropy', 'vowel-cons', 'ngrams'], \n class_names=['legit', 'isDGA'], \n discretize_continuous=False)", "_____no_output_____" ], [ "exp = explainer.explain_instance(feature_matrix_test.iloc[5], \n random_forest_clf.predict_proba, \n num_features=5)", "_____no_output_____" ], [ "exp.show_in_notebook(show_table=True, show_all=True)", "_____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", "markdown", "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code" ] ]

[ "WTFPL" ]

[ "WTFPL" ]

[ "WTFPL" ]

[ [ [ "# k-Nearest Neighbor (kNN) exercise\n\n*Complete and hand in this completed worksheet (including its outputs and any supporting code outside of the worksheet) with your assignment submission. For more details see the [assignments page](http://vision.stanford.edu/teaching/cs231n/assignments.html) on the course website.*\n\nThe kNN classifier consists of two stages:\n\n- During training, the classifier takes the training data and simply remembers it\n- During testing, kNN classifies every test image by comparing to all training images and transfering the labels of the k most similar training examples\n- The value of k is cross-validated\n\nIn this exercise you will implement these steps and understand the basic Image Classification pipeline, cross-validation, and gain proficiency in writing efficient, vectorized code.", "_____no_output_____" ] ], [ [ "# Run some setup code for this notebook.\n\nimport random\nimport numpy as np\nfrom cs231n.data_utils import load_CIFAR10\nimport matplotlib.pyplot as plt\n\n# This is a bit of magic to make matplotlib figures appear inline in the notebook\n# rather than in a new window.\n%matplotlib inline\nplt.rcParams['figure.figsize'] = (10.0, 8.0) # set default size of plots\nplt.rcParams['image.interpolation'] = 'nearest'\nplt.rcParams['image.cmap'] = 'gray'\n\n# Some more magic so that the notebook will reload external python modules;\n# see http://stackoverflow.com/questions/1907993/autoreload-of-modules-in-ipython\n%load_ext autoreload\n%autoreload 2", "_____no_output_____" ], [ "# Load the raw CIFAR-10 data.\ncifar10_dir = 'cs231n/datasets/cifar-10-batches-py'\nX_train, y_train, X_test, y_test = load_CIFAR10(cifar10_dir)\n\n# As a sanity check, we print out the size of the training and test data.\nprint 'Training data shape: ', X_train.shape\nprint 'Training labels shape: ', y_train.shape\nprint 'Test data shape: ', X_test.shape\nprint 'Test labels shape: ', y_test.shape", "Training data shape: (50000, 32, 32, 3)\nTraining labels shape: (50000,)\nTest data shape: (10000, 32, 32, 3)\nTest labels shape: (10000,)\n" ], [ "# Visualize some examples from the dataset.\n# We show a few examples of training images from each class.\nclasses = ['plane', 'car', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck']\nnum_classes = len(classes)\nsamples_per_class = 7\nfor y, cls in enumerate(classes):\n idxs = np.flatnonzero(y_train == y)\n idxs = np.random.choice(idxs, samples_per_class, replace=False)\n for i, idx in enumerate(idxs):\n plt_idx = i * num_classes + y + 1\n plt.subplot(samples_per_class, num_classes, plt_idx)\n plt.imshow(X_train[idx].astype('uint8'))\n plt.axis('off')\n if i == 0:\n plt.title(cls)\nplt.show()", "_____no_output_____" ], [ "# Subsample the data for more efficient code execution in this exercise\nnum_training = 5000\nmask = range(num_training)\nX_train = X_train[mask]\ny_train = y_train[mask]\n\nnum_test = 500\nmask = range(num_test)\nX_test = X_test[mask]\ny_test = y_test[mask]", "_____no_output_____" ], [ "X_train.shape, X_test.shape", "_____no_output_____" ], [ "# Reshape the image data into rows\nX_train = np.reshape(X_train, (X_train.shape[0], -1))\nX_test = np.reshape(X_test, (X_test.shape[0], -1))\nprint X_train.shape, X_test.shape", "(5000, 3072) (500, 3072)\n" ], [ "from cs231n.classifiers import KNearestNeighbor\n\n# Create a kNN classifier instance. \n# Remember that training a kNN classifier is a noop: \n# the Classifier simply remembers the data and does no further processing \nclassifier = KNearestNeighbor()\nclassifier.train(X_train, y_train)", "_____no_output_____" ] ], [ [ "We would now like to classify the test data with the kNN classifier. Recall that we can break down this process into two steps: \n\n1. First we must compute the distances between all test examples and all train examples. \n2. Given these distances, for each test example we find the k nearest examples and have them vote for the label\n\nLets begin with computing the distance matrix between all training and test examples. For example, if there are **Ntr** training examples and **Nte** test examples, this stage should result in a **Nte x Ntr** matrix where each element (i,j) is the distance between the i-th test and j-th train example.\n\nFirst, open `cs231n/classifiers/k_nearest_neighbor.py` and implement the function `compute_distances_two_loops` that uses a (very inefficient) double loop over all pairs of (test, train) examples and computes the distance matrix one element at a time.", "_____no_output_____" ] ], [ [ "# Open cs231n/classifiers/k_nearest_neighbor.py and implement\n# compute_distances_two_loops.\n\n# Test your implementation:\ndists = classifier.compute_distances_two_loops(X_test)\nprint dists.shape", "(500, 5000)\n" ], [ "# We can visualize the distance matrix: each row is a single test example and\n# its distances to training examples\nplt.imshow(dists, interpolation='none')\nplt.show()", "_____no_output_____" ] ], [ [ "**Inline Question #1:** Notice the structured patterns in the distance matrix, where some rows or columns are visible brighter. (Note that with the default color scheme black indicates low distances while white indicates high distances.)\n\n- What in the data is the cause behind the distinctly bright rows?\n- What causes the columns?", "_____no_output_____" ], [ "**Your Answer**:\n\n- The test image is too bright or too dark\n- The training data is too bright or too dark\n\n", "_____no_output_____" ] ], [ [ "# Now implement the function predict_labels and run the code below:\n# We use k = 1 (which is Nearest Neighbor).\ny_test_pred = classifier.predict_labels(dists, k=1)\n\n# Compute and print the fraction of correctly predicted examples\nnum_correct = np.sum(y_test_pred == y_test)\naccuracy = float(num_correct) / num_test\nprint 'Got %d / %d correct => accuracy: %f' % (num_correct, num_test, accuracy)", "Got 137 / 500 correct => accuracy: 0.274000\n" ] ], [ [ "You should expect to see approximately `27%` accuracy. Now lets try out a larger `k`, say `k = 5`:", "_____no_output_____" ] ], [ [ "y_test_pred = classifier.predict_labels(dists, k=5)\nnum_correct = np.sum(y_test_pred == y_test)\naccuracy = float(num_correct) / num_test\nprint 'Got %d / %d correct => accuracy: %f' % (num_correct, num_test, accuracy)", "Got 139 / 500 correct => accuracy: 0.278000\n" ] ], [ [ "You should expect to see a slightly better performance than with `k = 1`.", "_____no_output_____" ] ], [ [ "# Now lets speed up distance matrix computation by using partial vectorization\n# with one loop. Implement the function compute_distances_one_loop and run the\n# code below:\ndists_one = classifier.compute_distances_one_loop(X_test)\n\n# To ensure that our vectorized implementation is correct, we make sure that it\n# agrees with the naive implementation. There are many ways to decide whether\n# two matrices are similar; one of the simplest is the Frobenius norm. In case\n# you haven't seen it before, the Frobenius norm of two matrices is the square\n# root of the squared sum of differences of all elements; in other words, reshape\n# the matrices into vectors and compute the Euclidean distance between them.\ndifference = np.linalg.norm(dists - dists_one, ord='fro')\nprint 'Difference was: %f' % (difference, )\nif difference < 0.001:\n print 'Good! The distance matrices are the same'\nelse:\n print 'Uh-oh! The distance matrices are different'", "Difference was: 0.000000\nGood! The distance matrices are the same\n" ], [ "# Now implement the fully vectorized version inside compute_distances_no_loops\n# and run the code\ndists_two = classifier.compute_distances_no_loops(X_test)\n\n# check that the distance matrix agrees with the one we computed before:\ndifference = np.linalg.norm(dists - dists_two, ord='fro')\nprint 'Difference was: %f' % (difference, )\nif difference < 0.001:\n print 'Good! The distance matrices are the same'\nelse:\n print 'Uh-oh! The distance matrices are different'", "Difference was: 0.000000\nGood! The distance matrices are the same\n" ], [ "# Let's compare how fast the implementations are\ndef time_function(f, *args):\n \"\"\"\n Call a function f with args and return the time (in seconds) that it took to execute.\n \"\"\"\n import time\n tic = time.time()\n f(*args)\n toc = time.time()\n return toc - tic\n\ntwo_loop_time = time_function(classifier.compute_distances_two_loops, X_test)\nprint 'Two loop version took %f seconds' % two_loop_time\n\none_loop_time = time_function(classifier.compute_distances_one_loop, X_test)\nprint 'One loop version took %f seconds' % one_loop_time\n\nno_loop_time = time_function(classifier.compute_distances_no_loops, X_test)\nprint 'No loop version took %f seconds' % no_loop_time\n\n# you should see significantly faster performance with the fully vectorized implementation", "Two loop version took 27.158314 seconds\nOne loop version took 40.179075 seconds\nNo loop version took 0.529196 seconds\n" ] ], [ [ "### Cross-validation\n\nWe have implemented the k-Nearest Neighbor classifier but we set the value k = 5 arbitrarily. We will now determine the best value of this hyperparameter with cross-validation.", "_____no_output_____" ] ], [ [ "num_folds = 5\nk_choices = [1, 3, 5, 8, 10, 12, 15, 20, 50, 100]\n\nX_train_folds = []\ny_train_folds = []\n################################################################################\n# TODO: #\n# Split up the training data into folds. After splitting, X_train_folds and #\n# y_train_folds should each be lists of length num_folds, where #\n# y_train_folds[i] is the label vector for the points in X_train_folds[i]. #\n# Hint: Look up the numpy array_split function. #\n################################################################################\nX_train_folds = np.split(X_train, num_folds)\ny_train_folds = np.split(y_train, num_folds)\n################################################################################\n# END OF YOUR CODE #\n################################################################################\n\n# A dictionary holding the accuracies for different values of k that we find\n# when running cross-validation. After running cross-validation,\n# k_to_accuracies[k] should be a list of length num_folds giving the different\n# accuracy values that we found when using that value of k.\nk_to_accuracies = { k: [] for k in k_choices }\n\n################################################################################\n# TODO: #\n# Perform k-fold cross validation to find the best value of k. For each #\n# possible value of k, run the k-nearest-neighbor algorithm num_folds times, #\n# where in each case you use all but one of the folds as training data and the #\n# last fold as a validation set. Store the accuracies for all fold and all #\n# values of k in the k_to_accuracies dictionary. #\n################################################################################\nfor i in range(num_folds):\n classifier = KNearestNeighbor()\n X_test = X_train_folds[i]\n y_test = y_train_folds[i]\n X_train = np.concatenate([ fold for j, fold in enumerate(X_train_folds) if j != i ])\n y_train = np.concatenate([ fold for j, fold in enumerate(y_train_folds) if j != i ])\n classifier.train(X_train, y_train)\n dists = classifier.compute_distances_no_loops(X_test)\n for k in k_choices:\n predict = classifier.predict_labels(dists, k=k)\n num_correct = np.sum(predict == y_test)\n accuracy = float(num_correct) / X_test.shape[0]\n k_to_accuracies[k] += [accuracy]\n \n################################################################################\n# END OF YOUR CODE #\n################################################################################\n\n# Print out the computed accuracies\nfor k in sorted(k_to_accuracies):\n for accuracy in k_to_accuracies[k]:\n print 'k = %d, accuracy = %f' % (k, accuracy)", "k = 1, accuracy = 0.248438\nk = 1, accuracy = 0.237500\nk = 1, accuracy = 0.229687\nk = 1, accuracy = 0.285938\nk = 1, accuracy = 0.270313\nk = 3, accuracy = 0.228125\nk = 3, accuracy = 0.232813\nk = 3, accuracy = 0.240625\nk = 3, accuracy = 0.251563\nk = 3, accuracy = 0.248438\nk = 5, accuracy = 0.234375\nk = 5, accuracy = 0.245312\nk = 5, accuracy = 0.257812\nk = 5, accuracy = 0.271875\nk = 5, accuracy = 0.253125\nk = 8, accuracy = 0.231250\nk = 8, accuracy = 0.248438\nk = 8, accuracy = 0.281250\nk = 8, accuracy = 0.271875\nk = 8, accuracy = 0.260937\nk = 10, accuracy = 0.231250\nk = 10, accuracy = 0.257812\nk = 10, accuracy = 0.275000\nk = 10, accuracy = 0.276562\nk = 10, accuracy = 0.253125\nk = 12, accuracy = 0.225000\nk = 12, accuracy = 0.259375\nk = 12, accuracy = 0.279687\nk = 12, accuracy = 0.275000\nk = 12, accuracy = 0.242188\nk = 15, accuracy = 0.234375\nk = 15, accuracy = 0.268750\nk = 15, accuracy = 0.284375\nk = 15, accuracy = 0.257812\nk = 15, accuracy = 0.253125\nk = 20, accuracy = 0.245312\nk = 20, accuracy = 0.265625\nk = 20, accuracy = 0.290625\nk = 20, accuracy = 0.284375\nk = 20, accuracy = 0.253125\nk = 50, accuracy = 0.235937\nk = 50, accuracy = 0.243750\nk = 50, accuracy = 0.271875\nk = 50, accuracy = 0.242188\nk = 50, accuracy = 0.254688\nk = 100, accuracy = 0.240625\nk = 100, accuracy = 0.239063\nk = 100, accuracy = 0.254688\nk = 100, accuracy = 0.246875\nk = 100, accuracy = 0.246875\n" ], [ "# plot the raw observations\nfor k in k_choices:\n accuracies = k_to_accuracies[k]\n plt.scatter([k] * len(accuracies), accuracies)\n\n# plot the trend line with error bars that correspond to standard deviation\naccuracies_mean = np.array([np.mean(v) for k,v in sorted(k_to_accuracies.items())])\naccuracies_std = np.array([np.std(v) for k,v in sorted(k_to_accuracies.items())])\nplt.errorbar(k_choices, accuracies_mean, yerr=accuracies_std)\nplt.title('Cross-validation on k')\nplt.xlabel('k')\nplt.ylabel('Cross-validation accuracy')\nplt.show()", "_____no_output_____" ], [ "# Based on the cross-validation results above, choose the best value for k, \n# retrain the classifier using all the training data, and test it on the test\n# data. You should be able to get above 28% accuracy on the test data.\nbest_k = 20\n\nclassifier = KNearestNeighbor()\nclassifier.train(X_train, y_train)\ny_test_pred = classifier.predict(X_test, k=best_k)\n\n# Compute and display the accuracy\nnum_correct = np.sum(y_test_pred == y_test)\naccuracy = float(num_correct) / num_test\nprint 'Got %d / %d correct => accuracy: %f' % (num_correct, num_test, accuracy)", "Got 162 / 500 correct => accuracy: 0.324000\n" ] ] ]

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

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

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "## Monte Carlo - Forecasting Stock Prices - Part I", "_____no_output_____" ], [ "*Suggested Answers follow (usually there are multiple ways to solve a problem in Python).*", "_____no_output_____" ], [ "Load the data for Microsoft (‘MSFT’) for the period ‘2000-1-1’ until today.", "_____no_output_____" ] ], [ [ "import numpy as np \nimport pandas as pd \nfrom pandas_datareader import data as wb \nimport matplotlib.pyplot as plt \nfrom scipy.stats import norm\n%matplotlib inline\n\ndata = pd.read_csv('D:/Python/MSFT_2000.csv', index_col = 'Date')", "_____no_output_____" ] ], [ [ "Use the .pct_change() method to obtain the log returns of Microsoft for the designated period.", "_____no_output_____" ] ], [ [ "log_returns = np.log(1 + data.pct_change())", "_____no_output_____" ], [ "log_returns.tail()", "_____no_output_____" ], [ "data.plot(figsize=(10, 6));", "_____no_output_____" ], [ "log_returns.plot(figsize = (10, 6))", "_____no_output_____" ] ], [ [ "Assign the mean value of the log returns to a variable, called “U”, and their variance to a variable, called “var”. ", "_____no_output_____" ] ], [ [ "u = log_returns.mean()\nu", "_____no_output_____" ], [ "var = log_returns.var()\nvar", "_____no_output_____" ] ], [ [ "Calculate the drift, using the following formula: \n\n$$\ndrift = u - \\frac{1}{2} \\cdot var\n$$", "_____no_output_____" ] ], [ [ "drift = u - (0.5 * var)\ndrift", "_____no_output_____" ] ], [ [ "Store the standard deviation of the log returns in a variable, called “stdev”.", "_____no_output_____" ] ], [ [ "stdev = log_returns.std()\nstdev", "_____no_output_____" ] ], [ [ "******", "_____no_output_____" ], [ "Repeat this exercise for any stock of interest to you. :)", "_____no_output_____" ] ] ]

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

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

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Siamese Neural Network Recommendation for Friends (for Website)", "_____no_output_____" ], [ "This notebook presents the final code that will be used for the Movinder [website](https://movinder.herokuapp.com/) when `Get recommendation with SiameseNN!` is selected by user.", "_____no_output_____" ] ], [ [ "import pandas as pd\nimport json\nimport datetime, time\nfrom sklearn.model_selection import train_test_split\nimport itertools\nimport os\nimport zipfile\nimport random\nimport numpy as np\n\nimport requests\nimport matplotlib.pyplot as plt\n\nimport scipy.sparse as sp\nfrom sklearn.metrics import roc_auc_score", "_____no_output_____" ] ], [ [ "---", "_____no_output_____" ], [ "## (1) Read data", "_____no_output_____" ] ], [ [ "movies = json.load(open('movies.json'))\nfriends = json.load(open('friends.json'))\nratings = json.load(open('ratings.json')) \nsoup_movie_features = sp.load_npz('soup_movie_features_11.npz')\nsoup_movie_features = soup_movie_features.toarray()", "_____no_output_____" ] ], [ [ "## (1.2) Simulate new friend's input", "_____no_output_____" ], [ "The new group of friends will need to provide information that will be later used for training the model and predicting the ratings they will give to other movies. The friends will have a new id `new_friend_id`. They will provide a rating specified in the dictionary with the following keys: `movie_id_ml` (id of the movie rated), `rating` (rating of that movie on the scale from 1 to 5), and `friend_id` that will be the friends id specified as `new_friend_id`. In addition to this rating information, the users will have to provide to the system the information that includes their average age in the group `friends_age` and gender `friends_gender`.", "_____no_output_____" ] ], [ [ "new_friend_id = len(friends)", "_____no_output_____" ], [ "new_ratings = [{'movie_id_ml': 302.0, 'rating': 4.0, 'friend_id': new_friend_id},\n {'movie_id_ml': 304.0, 'rating': 4.0, 'friend_id': new_friend_id},\n {'movie_id_ml': 307.0, 'rating': 4.0, 'friend_id': new_friend_id}]\nnew_ratings", "_____no_output_____" ], [ "new_friend = {'friend_id': new_friend_id, 'friends_age': 25.5, 'friends_gender': 0.375}\nnew_friend", "_____no_output_____" ], [ "# extend the existing data with this new information\nfriends.append(new_friend)\nratings.extend(new_ratings)", "_____no_output_____" ] ], [ [ "---", "_____no_output_____" ], [ "## (2) Train the LightFM Model", "_____no_output_____" ], [ "We will be using the [LightFM](http://lyst.github.io/lightfm/docs/index.html) implementation of SiameseNN to train our model using the user and item (i.e. movie) features. First, we create `scipy.sparse` matrices from raw data and they can be used to fit the LightFM model.", "_____no_output_____" ] ], [ [ "from lightfm.data import Dataset\nfrom lightfm import LightFM\nfrom lightfm.evaluation import precision_at_k\nfrom lightfm.evaluation import auc_score", "_____no_output_____" ] ], [ [ "## (2.1) Build ID mappings", "_____no_output_____" ], [ "We create a mapping between the user and item ids from our input data to indices that will be internally used by this model. This needs to be done since the LightFM works with user and items ids that are consecutive non-negative integers. Using `dataset.fit` we assign internal numerical id to every user and item we passed in.", "_____no_output_____" ] ], [ [ "dataset = Dataset()\n\nitem_str_for_eval = \"x['title'],x['release'], x['unknown'], x['action'], x['adventure'],x['animation'], x['childrens'], x['comedy'], x['crime'], x['documentary'], x['drama'], x['fantasy'], x['noir'], x['horror'], x['musical'],x['mystery'], x['romance'], x['scifi'], x['thriller'], x['war'], x['western'], *soup_movie_features[x['soup_id']]\"\nfriend_str_for_eval = \"x['friends_age'], x['friends_gender']\"\n\n", "_____no_output_____" ], [ "dataset.fit(users=(int(x['friend_id']) for x in friends),\n items=(int(x['movie_id_ml']) for x in movies),\n item_features=(eval(\"(\"+item_str_for_eval+\")\") for x in movies),\n user_features=((eval(friend_str_for_eval)) for x in friends))\n\nnum_friends, num_items = dataset.interactions_shape()\nprint(f'Mappings - Num friends: {num_friends}, num_items {num_items}.')", "Mappings - Num friends: 192, num_items 1251.\n" ] ], [ [ "## (2.2) Build the interactions and feature matrices", "_____no_output_____" ], [ "The `interactions` matrix contains interactions between `friend_id` and `movie_id_ml`. It puts 1 if friends `friend_id` rated movie `movie_id_ml`, and 0 otherwise.", "_____no_output_____" ] ], [ [ "(interactions, weights) = dataset.build_interactions(((int(x['friend_id']), int(x['movie_id_ml']))\n for x in ratings))\n\nprint(repr(interactions))", "<192x1251 sparse matrix of type '<class 'numpy.int32'>'\n\twith 59123 stored elements in COOrdinate format>\n" ] ], [ [ "The `item_features` is also a sparse matrix that contains movie ids with their corresponding features. In the item features, we include the following features: movie title, when it was released, all genres it belongs to, and vectorized representation of movie keywords, cast members, and countries it was released in.", "_____no_output_____" ] ], [ [ "item_features = dataset.build_item_features(((x['movie_id_ml'], \n [eval(\"(\"+item_str_for_eval+\")\")]) for x in movies) )\nprint(repr(item_features))", "<1251x2487 sparse matrix of type '<class 'numpy.float32'>'\n\twith 2502 stored elements in Compressed Sparse Row format>\n" ] ], [ [ "The `user_features` is also a sparse matrix that contains movie ids with their corresponding features. The user features include their age, and gender.", "_____no_output_____" ] ], [ [ "user_features = dataset.build_user_features(((x['friend_id'], \n [eval(friend_str_for_eval)]) for x in friends) )\nprint(repr(user_features))", "<192x342 sparse matrix of type '<class 'numpy.float32'>'\n\twith 384 stored elements in Compressed Sparse Row format>\n" ] ], [ [ "## (2.3) Building a model", "_____no_output_____" ], [ "After some hyperparameters tuning, we end up to having the best model performance with the following values:\n\n- Epocks = 150\n- Learning rate = 0.015\n- Max sampled = 11\n- Loss type = WARP\n\nReferences:\n- The WARP (Weighted Approximate-Rank Pairwise) lso for implicit feedback learning-rank. Originally implemented in [WSABIE paper](http://www.thespermwhale.com/jaseweston/papers/wsabie-ijcai.pdf).\n- Extension to apply to recommendation settings in the 2013 k-order statistic loss [paper](http://www.ee.columbia.edu/~ronw/pubs/recsys2013-kaos.pdf) in the form of the k-OS WARP loss, also implemented in LightFM.", "_____no_output_____" ] ], [ [ "epochs = 150\nlr = 0.015\nmax_sampled = 11\n\nloss_type = \"warp\" # \"bpr\"\n\n\nmodel = LightFM(learning_rate=lr, loss=loss_type, max_sampled=max_sampled)\n\nmodel.fit_partial(interactions, epochs=epochs, user_features=user_features, item_features=item_features)\ntrain_precision = precision_at_k(model, interactions, k=10, user_features=user_features, item_features=item_features).mean()\n\ntrain_auc = auc_score(model, interactions, user_features=user_features, item_features=item_features).mean()\n\nprint(f'Precision: {train_precision}, AUC: {train_auc}')\n", "Precision: 0.9588541984558105, AUC: 0.9013209342956543\n" ], [ "def predict_top_k_movies(model, friends_id, k):\n n_users, n_movies = train.shape\n if use_features:\n prediction = model.predict(friends_id, np.arange(n_movies), user_features=friends_features, item_features=item_features)#predict(model, user_id, np.arange(n_movies))\n else:\n prediction = model.predict(friends_id, np.arange(n_movies))#predict(model, user_id, np.arange(n_movies))\n \n movie_ids = np.arange(train.shape[1])\n return movie_ids[np.argsort(-prediction)][:k]", "_____no_output_____" ], [ "dfm = pd.DataFrame(movies)\ndfm = dfm.sort_values(by=\"movie_id_ml\")", "_____no_output_____" ], [ "k = 10\nfriends_id = new_friend_id\nmovie_ids = np.array(dfm.movie_id_ml.unique())#np.array(list(df_movies.movie_id_ml.unique())) #np.arange(interactions.shape[1])\nprint(movie_ids.shape)\n\nn_users, n_items = interactions.shape\n\nscores = model.predict(friends_id, np.arange(n_items), user_features=user_features, item_features=item_features)\n# scores = model.predict(friends_id, np.arange(n_items))\n\nknown_positives = movie_ids[interactions.tocsr()[friends_id].indices]\ntop_items = movie_ids[np.argsort(-scores)]\n\nprint(f\"Friends {friends_id}\")\nprint(\" Known positives:\")\n\nfor x in known_positives[:k]:\n print(f\" {x} | {dfm[dfm.movie_id_ml==x]['title'].iloc[0]}\" )\n \nprint(\" Recommended:\")\nfor x in top_items[:k]:\n print(f\" {x} | {dfm[dfm.movie_id_ml==x]['title'].iloc[0]}\" )", "(1251,)\nFriends 191\n Known positives:\n 301 | in & out\n 302 | l.a. confidential\n 307 | the devil's advocate\n Recommended:\n 48 | hoop dreams\n 292 | rosewood\n 255 | my best friend's wedding\n 286 | the english patient\n 284 | tin cup\n 299 | hoodlum\n 125 | phenomenon\n 1 | toy story\n 315 | apt pupil\n 7 | twelve monkeys\n" ] ], [ [ "This is an example of recommended movies output that will be used in the website to give users a movie recommendation based on the information they supplied to the model.\n\nMovinder website: [https://movinder.herokuapp.com/](https://movinder.herokuapp.com/).", "_____no_output_____" ] ] ]

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

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "## Searching for bouts for a day of ephys recording\n- microphone wav file is first exported in sglx_pipe-dev-sort-bouts-s_b1253_21-20210614\n- bouts are extracted in searchbout_s_b1253_21-ephys", "_____no_output_____" ] ], [ [ "import os\nimport glob\nimport socket\nimport logging\nimport pickle\nimport numpy as np\nimport pandas as pd\nfrom scipy.io import wavfile\nfrom scipy import signal\n\n### Fuck matplotlib, I'm using poltly now\nfrom plotly.subplots import make_subplots\nimport plotly.graph_objects as go\n\nfrom importlib import reload\n\nlogger = logging.getLogger()\nhandler = logging.StreamHandler()\nformatter = logging.Formatter(\n '%(asctime)s %(name)-12s %(levelname)-8s %(message)s')\nhandler.setFormatter(formatter)\nlogger.addHandler(handler)\nlogger.setLevel(logging.INFO)\n\nlogger.info('Running on {}'.format(socket.gethostname()))", "2021-08-28 00:02:25,421 root INFO Running on pakhi\n" ], [ "from ceciestunepipe.file import bcistructure as et\nfrom ceciestunepipe.util.sound import boutsearch as bs", "_____no_output_____" ] ], [ [ "### Get the file locations for a session (day) of recordings", "_____no_output_____" ] ], [ [ "reload(et)\nsess_par = {'bird': 's_b1253_21',\n 'sess': '2021-07-18',\n 'sort': 2}\n\n\nexp_struct = et.get_exp_struct(sess_par['bird'], sess_par['sess'], ephys_software='sglx')\n\nraw_folder = exp_struct['folders']['sglx']\n\nderived_folder = exp_struct['folders']['derived']\n\nbouts_folder = os.path.join(derived_folder, 'bouts_ceciestunepipe')\n\nsess_bouts_file = os.path.join(bouts_folder, 'bout_sess_auto.pickle')\nsess_bouts_curated_file = os.path.join(bouts_folder, 'bout_curated.pickle')\n\nos.makedirs(bouts_folder, exist_ok=True)", "_____no_output_____" ], [ "exp_struct['folders']", "_____no_output_____" ] ], [ [ "### load concatenated the files of the session", "_____no_output_____" ] ], [ [ "def read_session_auto_bouts(exp_struct):\n # list all files of the session\n # read into list of pandas dataframes and concatenate\n # read the search parameters of the first session\n # return the big pd and the search params\n derived_folder = exp_struct['folders']['derived']\n bout_pd_files = et.get_sgl_files_epochs(derived_folder, file_filter='bout_auto.pickle')\n search_params_files = et.get_sgl_files_epochs(derived_folder, file_filter='bout_search_params.pickle')\n print(bout_pd_files)\n hparams=None\n with open(search_params_files[0], 'rb') as fh:\n hparams = pickle.load(fh)\n \n bout_pd = pd.concat([pd.read_pickle(p) for p in bout_pd_files[:]])\n \n return bout_pd, hparams\n\nbout_pd, hparams = read_session_auto_bouts(exp_struct)", "['/mnt/sphere/speech_bci/derived_data/s_b1253_21/2021-07-18/sglx/0610_g0/bout_auto.pickle', '/mnt/sphere/speech_bci/derived_data/s_b1253_21/2021-07-18/sglx/1615_g0/bout_auto.pickle']\n" ], [ "bout_pd['file'].values", "_____no_output_____" ] ], [ [ "###### if it wasnt saved (which is a bad mistake), read the sampling rate from the first file in the session", "_____no_output_____" ] ], [ [ "def sample_rate_from_wav(wav_path):\n x, sample_rate = wavfile.read(wav_path)\n return sample_rate\n\nif hparams['sample_rate'] is None:\n one_wav_path = bpd.loc[0, 'file']\n logger.info('Sample rate not saved in parameters dict, searching it in ' + one_wav_path)\n hparams['sample_rate'] = sample_rate_from_wav(one_wav_path)", "_____no_output_____" ], [ "def cleanup(bout_pd: pd.DataFrame):\n ## check for empty waveforms (how woudld THAT happen???)\n bout_pd['valid_waveform'] = bout_pd['waveform'].apply(lambda x: (False if x.size==0 else True))\n \n # valid is & of all the validated criteria\n bout_pd['valid'] = bout_pd['valid_waveform']\n \n ## fill in the epoch\n bout_pd['epoch'] = bout_pd['file'].apply(lambda x: et.split_path(x)[-2])\n \n # drop not valid and reset index\n bout_pd.drop(bout_pd[bout_pd['valid']==False].index, inplace=True)\n bout_pd.reset_index(drop=True, inplace=True)\n \n # set all to 'confusing' (unchecked) and 'bout_check' false (not a bout)\n bout_pd['confusing'] = True\n bout_pd['bout_check'] = False\n\ncleanup(bout_pd)", "2021-08-28 00:04:37,654 numexpr.utils INFO Note: NumExpr detected 32 cores but \"NUMEXPR_MAX_THREADS\" not set, so enforcing safe limit of 8.\n2021-08-28 00:04:37,655 numexpr.utils INFO NumExpr defaulting to 8 threads.\n" ], [ "bout_pd", "_____no_output_____" ], [ "reload(et)", "_____no_output_____" ] ], [ [ "#### compute the spectrograms", "_____no_output_____" ] ], [ [ "bout_pd['spectrogram'] = bout_pd['waveform'].apply(lambda x: bs.gimmepower(x, hparams)[2])\nlogger.info('saving bout pandas with spectrogram to ' + sess_bouts_file)\nbout_pd.to_pickle(sess_bouts_file)", "2021-08-28 00:05:31,806 root INFO saving bout pandas with spectrogram to /mnt/sphere/speech_bci/derived_data/s_b1253_21/2021-07-18/sglx/bouts_ceciestunepipe/bout_auto.pickle\n" ], [ "bout_pd.head(2)", "_____no_output_____" ], [ "bout_pd['file'][0]", "_____no_output_____" ] ], [ [ "## inspect the bouts and curate them", "_____no_output_____" ], [ "#### visualize one bout", "_____no_output_____" ] ], [ [ "bout_pd.iloc[0]", "_____no_output_____" ], [ "import plotly.express as px\nimport plotly.graph_objects as go\n\nfrom ipywidgets import widgets", "_____no_output_____" ], [ "def viz_one_bout(df: pd.Series, sub_sample=1):\n # get the power and the spectrogram\n sxx = df['spectrogram'][:, ::sub_sample]\n x = df['waveform'][::sub_sample]\n \n # the trace\n tr_waveform = go.Scatter(y=x)\n figwidg_waveform = go.FigureWidget(data=[tr_waveform],\n layout= {'height': 300,'width':1000})\n\n # the spectrogram\n fig_spectrogram = px.imshow(sxx, \n labels={}, \n color_continuous_scale='Inferno',\n aspect='auto')\n\n fig_spectrogram.update_layout(width=1000, height=300, coloraxis_showscale=False)\n fig_spectrogram.update_xaxes(showticklabels=False)\n fig_spectrogram.update_yaxes(showticklabels=False)\n \n \n figwidg_spectrogram = go.FigureWidget(fig_spectrogram)\n \n display(widgets.VBox([figwidg_waveform,\n figwidg_spectrogram]))\n \n\nviz_one_bout(bout_pd.iloc[24])", "_____no_output_____" ], [ "bout_pd.head(2)", "_____no_output_____" ] ], [ [ "## use it in a widget\n", "_____no_output_____" ], [ "#### add a 'confusing' label, for not/sure/mixed.\nwe want to avoid having things we are not sure of in the training dataset", "_____no_output_____" ] ], [ [ "bout_pd.reset_index(drop=True, inplace=True)", "_____no_output_____" ], [ "## Set confusing by default, will only be False once asserted bout/or not\nbout_pd['confusing'] = True\nbout_pd['bout_check'] = False", "_____no_output_____" ], [ "### Create a counter object (count goes 1:1 to DataFrame index)\nfrom traitlets import CInt, link\n\nclass Counter(widgets.DOMWidget):\n value = CInt(0)\n value.tag(sync=True)", "_____no_output_____" ], [ "class VizBout():\n def __init__(self, hparams, bouts_pd):\n self.bout = None\n self.bouts_pd = bouts_pd\n self.bout_series = None\n self.is_bout = None\n self.is_confusing = None\n \n self.bout_counter = None\n self.bout_id = None\n \n self.buttons = {}\n self.m_pick = None\n \n \n self.fig_waveform = None\n self.fig_spectrogram = None\n \n self.figwidg_waveform = None\n self.figwidg_spectrogram = None\n \n self.fig_width = 2\n self.sub_sample = 10\n \n self.x = None\n self.sxx = None\n self.tr_waveform = None\n \n self.s_f = hparams['sample_rate']\n \n self.init_fig()\n self.init_widget()\n self.show()\n \n def init_fig(self):\n # the trace\n self.tr_waveform = go.Scatter(y=np.zeros(500))\n self.figwidg_waveform = go.FigureWidget(data=[self.tr_waveform],\n layout={'width': 1000, 'height':300})\n \n # the spectrogram\n self.fig_spectrogram = px.imshow(np.random.rand(500, 500), \n labels={}, \n color_continuous_scale='Inferno',\n aspect='auto')\n \n self.fig_spectrogram.update_layout(width=1000, height=300, coloraxis_showscale=False)\n self.fig_spectrogram.update_xaxes(showticklabels=False)\n self.fig_spectrogram.update_yaxes(showticklabels=False)\n self.figwidg_spectrogram = go.FigureWidget(self.fig_spectrogram)\n \n \n def init_widget(self):\n # declare elements\n # lay them out\n #\n \n self.bout_counter = Counter()\n self.is_bout = widgets.Checkbox(description='is bout')\n self.is_confusing = widgets.Checkbox(description='Not sure or mixed')\n \n self.buttons['Next'] = widgets.Button(description=\"Next\", button_style='info',\n icon='plus') \n self.buttons['Prev'] = widgets.Button(description=\"Prev\", button_style='warning',\n icon='minus')\n self.buttons['Check'] = widgets.Button(description=\"Check\", button_style='success', \n icon='check')\n self.buttons['Uncheck'] = widgets.Button(description=\"Uncheck\", button_style='danger',\n icon='wrong')\n \n [b.on_click(self.button_click) for b in self.buttons.values()]\n \n left_box = widgets.VBox([self.buttons['Prev'], self.buttons['Uncheck']])\n right_box = widgets.VBox([self.buttons['Next'], self.buttons['Check']])\n button_box = widgets.HBox([left_box, right_box])\n\n self.m_pick = widgets.IntSlider(value=0, min=0, max=self.bouts_pd.index.size-1,step=1, \n description=\"Bout candidate index\")\n \n \n control_box = widgets.HBox([button_box,\n widgets.VBox([self.is_bout, self.is_confusing]),\n self.m_pick])\n \n link((self.m_pick, 'value'), (self.bout_counter, 'value'))\n\n self.update_bout()\n \n self.is_bout.observe(self.bout_checked, names='value')\n self.is_confusing.observe(self.confusing_checked, names='value')\n self.m_pick.observe(self.slider_change, names='value')\n \n all_containers = widgets.VBox([control_box, \n self.figwidg_waveform, self.figwidg_spectrogram])\n display(all_containers)\n# display(button_box)\n# display(self.m_pick)\n# display(self.is_bout)\n# display(self.fig)\n \n def button_click(self, button): \n self.bout_id = self.bout_counter.value\n curr_bout = self.bout_counter\n \n if button.description == 'Next':\n curr_bout.value += 1\n elif button.description == 'Prev':\n curr_bout.value -= 1\n elif button.description == 'Check':\n self.bouts_pd.loc[self.bout_id, 'bout_check'] = True\n self.bouts_pd.loc[self.bout_id, 'confusing'] = False\n curr_bout.value += 1\n elif button.description == 'Uncheck':\n self.bouts_pd.loc[self.bout_id, 'bout_check'] = False\n self.bouts_pd.loc[self.bout_id, 'confusing'] = False\n curr_bout.value += 1\n \n # handle the edges of the counter\n if curr_bout.value > self.m_pick.max:\n curr_bout.value = 0\n \n if curr_bout.value < self.m_pick.min:\n curr_bout.value = self.m_pick.max\n \n def slider_change(self, change):\n #logger.info('slider changed')\n #self.bout_counter = change.new\n #clear_output(True)\n self.update_bout()\n self.show()\n \n def bout_checked(self, bc):\n# print \"bout checked\"\n# print bc['new']\n# print self.motiff\n self.bouts_pd.loc[self.bout_id, 'bout_check'] = bc['new']\n \n def confusing_checked(self, bc):\n# print \"bout checked\"\n# print bc['new']\n# print self.motiff\n self.bouts_pd.loc[self.bout_id, 'confusing'] = bc['new']\n \n def update_bout(self):\n self.bout_id = self.bout_counter.value\n self.bout_series = self.bouts_pd.iloc[self.bout_id]\n \n self.is_bout.value = bool(self.bout_series['bout_check'])\n self.is_confusing.value = bool(self.bout_series['confusing'])\n \n self.x = self.bout_series['waveform'][::self.sub_sample]\n self.sxx = self.bout_series['spectrogram'][::self.sub_sample]\n \n def show(self):\n #self.fig.clf()\n #self.init_fig()\n # update\n# self.update_bout()\n #plot\n #logger.info('showing')\n \n # Show the figures\n with self.figwidg_waveform.batch_update():\n self.figwidg_waveform.data[0].y = self.x\n self.figwidg_waveform.data[0].x = np.arange(self.x.size) * self.sub_sample / self.s_f \n \n with self.figwidg_spectrogram.batch_update():\n self.figwidg_spectrogram.data[0].z = np.sqrt(self.sxx[::-1])\n \n \n\nviz_bout = VizBout(hparams, bout_pd)", "_____no_output_____" ], [ "np.where(viz_bout.bouts_pd['bout_check']==True)[0].size", "_____no_output_____" ] ], [ [ "### save it", "_____no_output_____" ] ], [ [ "hparams", "_____no_output_____" ], [ "### get the curated file path\n##save to the curated file path\nviz_bout.bouts_pd.to_pickle(sess_bouts_curated_file)\nlogger.info('saved curated bout pandas to pickle {}'.format(sess_bouts_curated_file))", "2021-08-28 00:18:56,404 root INFO saved curated bout pandas to pickle /mnt/sphere/speech_bci/derived_data/s_b1253_21/2021-07-18/sglx/bouts_ceciestunepipe/bout_curated.pickle\n" ], [ "viz_bout.bouts_pd['file'][0]", "_____no_output_____" ], [ "viz_bout.bouts_pd.head(5)", "_____no_output_____" ] ] ]

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

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

[ "BSD-2-Clause" ]

[ "BSD-2-Clause" ]

[ "BSD-2-Clause" ]

[ [ [ "empty" ] ] ]

[ "empty" ]

[ [ "empty" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Simple CNN on dataset", "_____no_output_____" ] ], [ [ "import numpy as np\nimport pandas as pd\nimport keras\nimport matplotlib.pyplot as plt\nimport os \nos.environ['CUDA_VISIBLE_DEVICES'] = '1'\n\nfrom keras.preprocessing.image import ImageDataGenerator\nfrom keras.models import Sequential, Model\nfrom keras.layers import Conv2D, MaxPooling2D\nfrom keras.layers import Activation, Dropout, Flatten, Dense, BatchNormalization, GlobalAveragePooling2D\nfrom keras import backend as K\nfrom keras.utils import multi_gpu_model\n", "_____no_output_____" ], [ "# dimensions of our images.\nimg_width, img_height = 224, 224\n\ntrain_data_dir = '../dataset/train'\nvalidation_data_dir = '../dataset/test'\nnb_train_samples = 194\nnb_validation_samples = 49\nepochs = 50\nbatch_size = 4\n\nif K.image_data_format() == 'channels_first':\n input_shape = (3, img_width, img_height)\nelse:\n input_shape = (img_width, img_height, 3)\n", "_____no_output_____" ], [ "from keras.applications.resnet import ResNet50\nimport numpy as np\n\nbase_model = ResNet50(weights='imagenet', input_shape=(img_width, img_height, 3), include_top = False, pooling='max')\n\n#adam = keras.optimizers.Adam(lr=1e-4)\n#model.compile(loss='binary_crossentropy',\n# optimizer=adam,\n# metrics=['accuracy'])\n\nx = base_model.output\n# x = Flatten()(x)\nx = Dense(512, activation='relu')(x)\nx = BatchNormalization()(x)\npredictions = Dense(1, activation='sigmoid')(x)\n\n# this is the model we will train\nmodel = Model(inputs=base_model.input, outputs=predictions)\n\n\nadam = keras.optimizers.Adam(lr=1e-4)\nmodel.compile(loss='binary_crossentropy',\n optimizer=adam,\n metrics=['accuracy'])\n\nmodel.summary()", "WARNING:tensorflow:From /usr/local/lib/python3.5/dist-packages/keras/backend/tensorflow_backend.py:2041: The name tf.nn.fused_batch_norm is deprecated. Please use tf.compat.v1.nn.fused_batch_norm instead.\n\nDownloading data from https://github.com/keras-team/keras-applications/releases/download/resnet/resnet50_weights_tf_dim_ordering_tf_kernels_notop.h5\n94773248/94765736 [==============================] - 11s 0us/step\nModel: \"model_3\"\n__________________________________________________________________________________________________\nLayer (type) Output Shape Param # Connected to \n==================================================================================================\ninput_3 (InputLayer) (None, 224, 224, 3) 0 \n__________________________________________________________________________________________________\nconv1_pad (ZeroPadding2D) (None, 230, 230, 3) 0 input_3[0][0] \n__________________________________________________________________________________________________\nconv1_conv (Conv2D) (None, 112, 112, 64) 9472 conv1_pad[0][0] \n__________________________________________________________________________________________________\nconv1_bn (BatchNormalization) (None, 112, 112, 64) 256 conv1_conv[0][0] \n__________________________________________________________________________________________________\nconv1_relu (Activation) (None, 112, 112, 64) 0 conv1_bn[0][0] \n__________________________________________________________________________________________________\npool1_pad (ZeroPadding2D) (None, 114, 114, 64) 0 conv1_relu[0][0] \n__________________________________________________________________________________________________\npool1_pool (MaxPooling2D) (None, 56, 56, 64) 0 pool1_pad[0][0] \n__________________________________________________________________________________________________\nconv2_block1_1_conv (Conv2D) (None, 56, 56, 64) 4160 pool1_pool[0][0] \n__________________________________________________________________________________________________\nconv2_block1_1_bn (BatchNormali (None, 56, 56, 64) 256 conv2_block1_1_conv[0][0] \n__________________________________________________________________________________________________\nconv2_block1_1_relu (Activation (None, 56, 56, 64) 0 conv2_block1_1_bn[0][0] \n__________________________________________________________________________________________________\nconv2_block1_2_conv (Conv2D) (None, 56, 56, 64) 36928 conv2_block1_1_relu[0][0] \n__________________________________________________________________________________________________\nconv2_block1_2_bn (BatchNormali (None, 56, 56, 64) 256 conv2_block1_2_conv[0][0] \n__________________________________________________________________________________________________\nconv2_block1_2_relu (Activation (None, 56, 56, 64) 0 conv2_block1_2_bn[0][0] \n__________________________________________________________________________________________________\nconv2_block1_0_conv (Conv2D) (None, 56, 56, 256) 16640 pool1_pool[0][0] \n__________________________________________________________________________________________________\nconv2_block1_3_conv (Conv2D) (None, 56, 56, 256) 16640 conv2_block1_2_relu[0][0] \n__________________________________________________________________________________________________\nconv2_block1_0_bn (BatchNormali (None, 56, 56, 256) 1024 conv2_block1_0_conv[0][0] \n__________________________________________________________________________________________________\nconv2_block1_3_bn (BatchNormali (None, 56, 56, 256) 1024 conv2_block1_3_conv[0][0] \n__________________________________________________________________________________________________\nconv2_block1_add (Add) (None, 56, 56, 256) 0 conv2_block1_0_bn[0][0] \n conv2_block1_3_bn[0][0] \n__________________________________________________________________________________________________\nconv2_block1_out (Activation) (None, 56, 56, 256) 0 conv2_block1_add[0][0] \n__________________________________________________________________________________________________\nconv2_block2_1_conv (Conv2D) (None, 56, 56, 64) 16448 conv2_block1_out[0][0] \n__________________________________________________________________________________________________\nconv2_block2_1_bn (BatchNormali (None, 56, 56, 64) 256 conv2_block2_1_conv[0][0] \n__________________________________________________________________________________________________\nconv2_block2_1_relu (Activation (None, 56, 56, 64) 0 conv2_block2_1_bn[0][0] \n__________________________________________________________________________________________________\nconv2_block2_2_conv (Conv2D) (None, 56, 56, 64) 36928 conv2_block2_1_relu[0][0] \n__________________________________________________________________________________________________\nconv2_block2_2_bn (BatchNormali (None, 56, 56, 64) 256 conv2_block2_2_conv[0][0] \n__________________________________________________________________________________________________\nconv2_block2_2_relu (Activation (None, 56, 56, 64) 0 conv2_block2_2_bn[0][0] \n__________________________________________________________________________________________________\nconv2_block2_3_conv (Conv2D) (None, 56, 56, 256) 16640 conv2_block2_2_relu[0][0] \n__________________________________________________________________________________________________\nconv2_block2_3_bn (BatchNormali (None, 56, 56, 256) 1024 conv2_block2_3_conv[0][0] \n__________________________________________________________________________________________________\nconv2_block2_add (Add) (None, 56, 56, 256) 0 conv2_block1_out[0][0] \n conv2_block2_3_bn[0][0] \n__________________________________________________________________________________________________\nconv2_block2_out (Activation) (None, 56, 56, 256) 0 conv2_block2_add[0][0] \n__________________________________________________________________________________________________\nconv2_block3_1_conv (Conv2D) (None, 56, 56, 64) 16448 conv2_block2_out[0][0] \n__________________________________________________________________________________________________\nconv2_block3_1_bn (BatchNormali (None, 56, 56, 64) 256 conv2_block3_1_conv[0][0] \n__________________________________________________________________________________________________\nconv2_block3_1_relu (Activation (None, 56, 56, 64) 0 conv2_block3_1_bn[0][0] \n__________________________________________________________________________________________________\nconv2_block3_2_conv (Conv2D) (None, 56, 56, 64) 36928 conv2_block3_1_relu[0][0] \n__________________________________________________________________________________________________\nconv2_block3_2_bn (BatchNormali (None, 56, 56, 64) 256 conv2_block3_2_conv[0][0] \n__________________________________________________________________________________________________\nconv2_block3_2_relu (Activation (None, 56, 56, 64) 0 conv2_block3_2_bn[0][0] \n__________________________________________________________________________________________________\nconv2_block3_3_conv (Conv2D) (None, 56, 56, 256) 16640 conv2_block3_2_relu[0][0] \n__________________________________________________________________________________________________\nconv2_block3_3_bn (BatchNormali (None, 56, 56, 256) 1024 conv2_block3_3_conv[0][0] \n__________________________________________________________________________________________________\nconv2_block3_add (Add) (None, 56, 56, 256) 0 conv2_block2_out[0][0] \n conv2_block3_3_bn[0][0] \n__________________________________________________________________________________________________\nconv2_block3_out (Activation) (None, 56, 56, 256) 0 conv2_block3_add[0][0] \n__________________________________________________________________________________________________\nconv3_block1_1_conv (Conv2D) (None, 28, 28, 128) 32896 conv2_block3_out[0][0] \n__________________________________________________________________________________________________\nconv3_block1_1_bn (BatchNormali (None, 28, 28, 128) 512 conv3_block1_1_conv[0][0] \n__________________________________________________________________________________________________\nconv3_block1_1_relu (Activation (None, 28, 28, 128) 0 conv3_block1_1_bn[0][0] \n__________________________________________________________________________________________________\nconv3_block1_2_conv (Conv2D) (None, 28, 28, 128) 147584 conv3_block1_1_relu[0][0] \n__________________________________________________________________________________________________\nconv3_block1_2_bn (BatchNormali (None, 28, 28, 128) 512 conv3_block1_2_conv[0][0] \n__________________________________________________________________________________________________\nconv3_block1_2_relu (Activation (None, 28, 28, 128) 0 conv3_block1_2_bn[0][0] \n__________________________________________________________________________________________________\nconv3_block1_0_conv (Conv2D) (None, 28, 28, 512) 131584 conv2_block3_out[0][0] \n__________________________________________________________________________________________________\nconv3_block1_3_conv (Conv2D) (None, 28, 28, 512) 66048 conv3_block1_2_relu[0][0] \n__________________________________________________________________________________________________\nconv3_block1_0_bn (BatchNormali (None, 28, 28, 512) 2048 conv3_block1_0_conv[0][0] \n__________________________________________________________________________________________________\nconv3_block1_3_bn (BatchNormali (None, 28, 28, 512) 2048 conv3_block1_3_conv[0][0] \n__________________________________________________________________________________________________\nconv3_block1_add (Add) (None, 28, 28, 512) 0 conv3_block1_0_bn[0][0] \n conv3_block1_3_bn[0][0] \n__________________________________________________________________________________________________\nconv3_block1_out (Activation) (None, 28, 28, 512) 0 conv3_block1_add[0][0] \n__________________________________________________________________________________________________\nconv3_block2_1_conv (Conv2D) (None, 28, 28, 128) 65664 conv3_block1_out[0][0] \n__________________________________________________________________________________________________\nconv3_block2_1_bn (BatchNormali (None, 28, 28, 128) 512 conv3_block2_1_conv[0][0] \n__________________________________________________________________________________________________\nconv3_block2_1_relu (Activation (None, 28, 28, 128) 0 conv3_block2_1_bn[0][0] \n__________________________________________________________________________________________________\nconv3_block2_2_conv (Conv2D) (None, 28, 28, 128) 147584 conv3_block2_1_relu[0][0] \n__________________________________________________________________________________________________\nconv3_block2_2_bn (BatchNormali (None, 28, 28, 128) 512 conv3_block2_2_conv[0][0] \n__________________________________________________________________________________________________\nconv3_block2_2_relu (Activation (None, 28, 28, 128) 0 conv3_block2_2_bn[0][0] \n__________________________________________________________________________________________________\nconv3_block2_3_conv (Conv2D) (None, 28, 28, 512) 66048 conv3_block2_2_relu[0][0] \n__________________________________________________________________________________________________\nconv3_block2_3_bn (BatchNormali (None, 28, 28, 512) 2048 conv3_block2_3_conv[0][0] \n__________________________________________________________________________________________________\nconv3_block2_add (Add) (None, 28, 28, 512) 0 conv3_block1_out[0][0] \n conv3_block2_3_bn[0][0] \n__________________________________________________________________________________________________\nconv3_block2_out (Activation) (None, 28, 28, 512) 0 conv3_block2_add[0][0] \n__________________________________________________________________________________________________\nconv3_block3_1_conv (Conv2D) (None, 28, 28, 128) 65664 conv3_block2_out[0][0] \n__________________________________________________________________________________________________\nconv3_block3_1_bn (BatchNormali (None, 28, 28, 128) 512 conv3_block3_1_conv[0][0] \n__________________________________________________________________________________________________\nconv3_block3_1_relu (Activation (None, 28, 28, 128) 0 conv3_block3_1_bn[0][0] \n__________________________________________________________________________________________________\nconv3_block3_2_conv (Conv2D) (None, 28, 28, 128) 147584 conv3_block3_1_relu[0][0] \n__________________________________________________________________________________________________\nconv3_block3_2_bn (BatchNormali (None, 28, 28, 128) 512 conv3_block3_2_conv[0][0] \n__________________________________________________________________________________________________\nconv3_block3_2_relu (Activation (None, 28, 28, 128) 0 conv3_block3_2_bn[0][0] \n__________________________________________________________________________________________________\nconv3_block3_3_conv (Conv2D) (None, 28, 28, 512) 66048 conv3_block3_2_relu[0][0] \n__________________________________________________________________________________________________\nconv3_block3_3_bn (BatchNormali (None, 28, 28, 512) 2048 conv3_block3_3_conv[0][0] \n__________________________________________________________________________________________________\nconv3_block3_add (Add) (None, 28, 28, 512) 0 conv3_block2_out[0][0] \n conv3_block3_3_bn[0][0] \n__________________________________________________________________________________________________\nconv3_block3_out (Activation) (None, 28, 28, 512) 0 conv3_block3_add[0][0] \n__________________________________________________________________________________________________\nconv3_block4_1_conv (Conv2D) (None, 28, 28, 128) 65664 conv3_block3_out[0][0] \n__________________________________________________________________________________________________\nconv3_block4_1_bn (BatchNormali (None, 28, 28, 128) 512 conv3_block4_1_conv[0][0] \n__________________________________________________________________________________________________\nconv3_block4_1_relu (Activation (None, 28, 28, 128) 0 conv3_block4_1_bn[0][0] \n__________________________________________________________________________________________________\nconv3_block4_2_conv (Conv2D) (None, 28, 28, 128) 147584 conv3_block4_1_relu[0][0] \n__________________________________________________________________________________________________\nconv3_block4_2_bn (BatchNormali (None, 28, 28, 128) 512 conv3_block4_2_conv[0][0] \n__________________________________________________________________________________________________\nconv3_block4_2_relu (Activation (None, 28, 28, 128) 0 conv3_block4_2_bn[0][0] \n__________________________________________________________________________________________________\nconv3_block4_3_conv (Conv2D) (None, 28, 28, 512) 66048 conv3_block4_2_relu[0][0] \n__________________________________________________________________________________________________\nconv3_block4_3_bn (BatchNormali (None, 28, 28, 512) 2048 conv3_block4_3_conv[0][0] \n__________________________________________________________________________________________________\nconv3_block4_add (Add) (None, 28, 28, 512) 0 conv3_block3_out[0][0] \n conv3_block4_3_bn[0][0] \n__________________________________________________________________________________________________\nconv3_block4_out (Activation) (None, 28, 28, 512) 0 conv3_block4_add[0][0] \n__________________________________________________________________________________________________\nconv4_block1_1_conv (Conv2D) (None, 14, 14, 256) 131328 conv3_block4_out[0][0] \n__________________________________________________________________________________________________\nconv4_block1_1_bn (BatchNormali (None, 14, 14, 256) 1024 conv4_block1_1_conv[0][0] \n__________________________________________________________________________________________________\nconv4_block1_1_relu (Activation (None, 14, 14, 256) 0 conv4_block1_1_bn[0][0] \n__________________________________________________________________________________________________\nconv4_block1_2_conv (Conv2D) (None, 14, 14, 256) 590080 conv4_block1_1_relu[0][0] \n__________________________________________________________________________________________________\nconv4_block1_2_bn (BatchNormali (None, 14, 14, 256) 1024 conv4_block1_2_conv[0][0] \n__________________________________________________________________________________________________\nconv4_block1_2_relu (Activation (None, 14, 14, 256) 0 conv4_block1_2_bn[0][0] \n__________________________________________________________________________________________________\nconv4_block1_0_conv (Conv2D) (None, 14, 14, 1024) 525312 conv3_block4_out[0][0] \n__________________________________________________________________________________________________\nconv4_block1_3_conv (Conv2D) (None, 14, 14, 1024) 263168 conv4_block1_2_relu[0][0] \n__________________________________________________________________________________________________\nconv4_block1_0_bn (BatchNormali (None, 14, 14, 1024) 4096 conv4_block1_0_conv[0][0] \n__________________________________________________________________________________________________\nconv4_block1_3_bn (BatchNormali (None, 14, 14, 1024) 4096 conv4_block1_3_conv[0][0] \n__________________________________________________________________________________________________\nconv4_block1_add (Add) (None, 14, 14, 1024) 0 conv4_block1_0_bn[0][0] \n conv4_block1_3_bn[0][0] \n__________________________________________________________________________________________________\nconv4_block1_out (Activation) (None, 14, 14, 1024) 0 conv4_block1_add[0][0] \n__________________________________________________________________________________________________\nconv4_block2_1_conv (Conv2D) (None, 14, 14, 256) 262400 conv4_block1_out[0][0] \n__________________________________________________________________________________________________\nconv4_block2_1_bn (BatchNormali (None, 14, 14, 256) 1024 conv4_block2_1_conv[0][0] \n__________________________________________________________________________________________________\nconv4_block2_1_relu (Activation (None, 14, 14, 256) 0 conv4_block2_1_bn[0][0] \n__________________________________________________________________________________________________\nconv4_block2_2_conv (Conv2D) (None, 14, 14, 256) 590080 conv4_block2_1_relu[0][0] \n__________________________________________________________________________________________________\nconv4_block2_2_bn (BatchNormali (None, 14, 14, 256) 1024 conv4_block2_2_conv[0][0] \n__________________________________________________________________________________________________\nconv4_block2_2_relu (Activation (None, 14, 14, 256) 0 conv4_block2_2_bn[0][0] \n__________________________________________________________________________________________________\nconv4_block2_3_conv (Conv2D) (None, 14, 14, 1024) 263168 conv4_block2_2_relu[0][0] \n__________________________________________________________________________________________________\nconv4_block2_3_bn (BatchNormali (None, 14, 14, 1024) 4096 conv4_block2_3_conv[0][0] \n__________________________________________________________________________________________________\nconv4_block2_add (Add) (None, 14, 14, 1024) 0 conv4_block1_out[0][0] \n conv4_block2_3_bn[0][0] \n__________________________________________________________________________________________________\nconv4_block2_out (Activation) (None, 14, 14, 1024) 0 conv4_block2_add[0][0] \n__________________________________________________________________________________________________\nconv4_block3_1_conv (Conv2D) (None, 14, 14, 256) 262400 conv4_block2_out[0][0] \n__________________________________________________________________________________________________\nconv4_block3_1_bn (BatchNormali (None, 14, 14, 256) 1024 conv4_block3_1_conv[0][0] \n__________________________________________________________________________________________________\nconv4_block3_1_relu (Activation (None, 14, 14, 256) 0 conv4_block3_1_bn[0][0] \n__________________________________________________________________________________________________\nconv4_block3_2_conv (Conv2D) (None, 14, 14, 256) 590080 conv4_block3_1_relu[0][0] \n__________________________________________________________________________________________________\nconv4_block3_2_bn (BatchNormali (None, 14, 14, 256) 1024 conv4_block3_2_conv[0][0] \n__________________________________________________________________________________________________\nconv4_block3_2_relu (Activation (None, 14, 14, 256) 0 conv4_block3_2_bn[0][0] \n__________________________________________________________________________________________________\nconv4_block3_3_conv (Conv2D) (None, 14, 14, 1024) 263168 conv4_block3_2_relu[0][0] \n__________________________________________________________________________________________________\nconv4_block3_3_bn (BatchNormali (None, 14, 14, 1024) 4096 conv4_block3_3_conv[0][0] \n__________________________________________________________________________________________________\nconv4_block3_add (Add) (None, 14, 14, 1024) 0 conv4_block2_out[0][0] \n conv4_block3_3_bn[0][0] \n__________________________________________________________________________________________________\nconv4_block3_out (Activation) (None, 14, 14, 1024) 0 conv4_block3_add[0][0] \n__________________________________________________________________________________________________\nconv4_block4_1_conv (Conv2D) (None, 14, 14, 256) 262400 conv4_block3_out[0][0] \n__________________________________________________________________________________________________\nconv4_block4_1_bn (BatchNormali (None, 14, 14, 256) 1024 conv4_block4_1_conv[0][0] \n__________________________________________________________________________________________________\nconv4_block4_1_relu (Activation (None, 14, 14, 256) 0 conv4_block4_1_bn[0][0] \n__________________________________________________________________________________________________\nconv4_block4_2_conv (Conv2D) (None, 14, 14, 256) 590080 conv4_block4_1_relu[0][0] \n__________________________________________________________________________________________________\nconv4_block4_2_bn (BatchNormali (None, 14, 14, 256) 1024 conv4_block4_2_conv[0][0] \n__________________________________________________________________________________________________\nconv4_block4_2_relu (Activation (None, 14, 14, 256) 0 conv4_block4_2_bn[0][0] \n__________________________________________________________________________________________________\nconv4_block4_3_conv (Conv2D) (None, 14, 14, 1024) 263168 conv4_block4_2_relu[0][0] \n__________________________________________________________________________________________________\nconv4_block4_3_bn (BatchNormali (None, 14, 14, 1024) 4096 conv4_block4_3_conv[0][0] \n__________________________________________________________________________________________________\nconv4_block4_add (Add) (None, 14, 14, 1024) 0 conv4_block3_out[0][0] \n conv4_block4_3_bn[0][0] \n__________________________________________________________________________________________________\nconv4_block4_out (Activation) (None, 14, 14, 1024) 0 conv4_block4_add[0][0] \n__________________________________________________________________________________________________\nconv4_block5_1_conv (Conv2D) (None, 14, 14, 256) 262400 conv4_block4_out[0][0] \n__________________________________________________________________________________________________\nconv4_block5_1_bn (BatchNormali (None, 14, 14, 256) 1024 conv4_block5_1_conv[0][0] \n__________________________________________________________________________________________________\nconv4_block5_1_relu (Activation (None, 14, 14, 256) 0 conv4_block5_1_bn[0][0] \n__________________________________________________________________________________________________\nconv4_block5_2_conv (Conv2D) (None, 14, 14, 256) 590080 conv4_block5_1_relu[0][0] \n__________________________________________________________________________________________________\nconv4_block5_2_bn (BatchNormali (None, 14, 14, 256) 1024 conv4_block5_2_conv[0][0] \n__________________________________________________________________________________________________\nconv4_block5_2_relu (Activation (None, 14, 14, 256) 0 conv4_block5_2_bn[0][0] \n__________________________________________________________________________________________________\nconv4_block5_3_conv (Conv2D) (None, 14, 14, 1024) 263168 conv4_block5_2_relu[0][0] \n__________________________________________________________________________________________________\nconv4_block5_3_bn (BatchNormali (None, 14, 14, 1024) 4096 conv4_block5_3_conv[0][0] \n__________________________________________________________________________________________________\nconv4_block5_add (Add) (None, 14, 14, 1024) 0 conv4_block4_out[0][0] \n conv4_block5_3_bn[0][0] \n__________________________________________________________________________________________________\nconv4_block5_out (Activation) (None, 14, 14, 1024) 0 conv4_block5_add[0][0] \n__________________________________________________________________________________________________\nconv4_block6_1_conv (Conv2D) (None, 14, 14, 256) 262400 conv4_block5_out[0][0] \n__________________________________________________________________________________________________\nconv4_block6_1_bn (BatchNormali (None, 14, 14, 256) 1024 conv4_block6_1_conv[0][0] \n__________________________________________________________________________________________________\nconv4_block6_1_relu (Activation (None, 14, 14, 256) 0 conv4_block6_1_bn[0][0] \n__________________________________________________________________________________________________\nconv4_block6_2_conv (Conv2D) (None, 14, 14, 256) 590080 conv4_block6_1_relu[0][0] \n__________________________________________________________________________________________________\nconv4_block6_2_bn (BatchNormali (None, 14, 14, 256) 1024 conv4_block6_2_conv[0][0] \n__________________________________________________________________________________________________\nconv4_block6_2_relu (Activation (None, 14, 14, 256) 0 conv4_block6_2_bn[0][0] \n__________________________________________________________________________________________________\nconv4_block6_3_conv (Conv2D) (None, 14, 14, 1024) 263168 conv4_block6_2_relu[0][0] \n__________________________________________________________________________________________________\nconv4_block6_3_bn (BatchNormali (None, 14, 14, 1024) 4096 conv4_block6_3_conv[0][0] \n__________________________________________________________________________________________________\nconv4_block6_add (Add) (None, 14, 14, 1024) 0 conv4_block5_out[0][0] \n conv4_block6_3_bn[0][0] \n__________________________________________________________________________________________________\nconv4_block6_out (Activation) (None, 14, 14, 1024) 0 conv4_block6_add[0][0] \n__________________________________________________________________________________________________\nconv5_block1_1_conv (Conv2D) (None, 7, 7, 512) 524800 conv4_block6_out[0][0] \n__________________________________________________________________________________________________\nconv5_block1_1_bn (BatchNormali (None, 7, 7, 512) 2048 conv5_block1_1_conv[0][0] \n__________________________________________________________________________________________________\nconv5_block1_1_relu (Activation (None, 7, 7, 512) 0 conv5_block1_1_bn[0][0] \n__________________________________________________________________________________________________\nconv5_block1_2_conv (Conv2D) (None, 7, 7, 512) 2359808 conv5_block1_1_relu[0][0] \n__________________________________________________________________________________________________\nconv5_block1_2_bn (BatchNormali (None, 7, 7, 512) 2048 conv5_block1_2_conv[0][0] \n__________________________________________________________________________________________________\nconv5_block1_2_relu (Activation (None, 7, 7, 512) 0 conv5_block1_2_bn[0][0] \n__________________________________________________________________________________________________\nconv5_block1_0_conv (Conv2D) (None, 7, 7, 2048) 2099200 conv4_block6_out[0][0] \n__________________________________________________________________________________________________\nconv5_block1_3_conv (Conv2D) (None, 7, 7, 2048) 1050624 conv5_block1_2_relu[0][0] \n__________________________________________________________________________________________________\nconv5_block1_0_bn (BatchNormali (None, 7, 7, 2048) 8192 conv5_block1_0_conv[0][0] \n__________________________________________________________________________________________________\nconv5_block1_3_bn (BatchNormali (None, 7, 7, 2048) 8192 conv5_block1_3_conv[0][0] \n__________________________________________________________________________________________________\nconv5_block1_add (Add) (None, 7, 7, 2048) 0 conv5_block1_0_bn[0][0] \n conv5_block1_3_bn[0][0] \n__________________________________________________________________________________________________\nconv5_block1_out (Activation) (None, 7, 7, 2048) 0 conv5_block1_add[0][0] \n__________________________________________________________________________________________________\nconv5_block2_1_conv (Conv2D) (None, 7, 7, 512) 1049088 conv5_block1_out[0][0] \n__________________________________________________________________________________________________\nconv5_block2_1_bn (BatchNormali (None, 7, 7, 512) 2048 conv5_block2_1_conv[0][0] \n__________________________________________________________________________________________________\nconv5_block2_1_relu (Activation (None, 7, 7, 512) 0 conv5_block2_1_bn[0][0] \n__________________________________________________________________________________________________\nconv5_block2_2_conv (Conv2D) (None, 7, 7, 512) 2359808 conv5_block2_1_relu[0][0] \n__________________________________________________________________________________________________\nconv5_block2_2_bn (BatchNormali (None, 7, 7, 512) 2048 conv5_block2_2_conv[0][0] \n__________________________________________________________________________________________________\nconv5_block2_2_relu (Activation (None, 7, 7, 512) 0 conv5_block2_2_bn[0][0] \n__________________________________________________________________________________________________\nconv5_block2_3_conv (Conv2D) (None, 7, 7, 2048) 1050624 conv5_block2_2_relu[0][0] \n__________________________________________________________________________________________________\nconv5_block2_3_bn (BatchNormali (None, 7, 7, 2048) 8192 conv5_block2_3_conv[0][0] \n__________________________________________________________________________________________________\nconv5_block2_add (Add) (None, 7, 7, 2048) 0 conv5_block1_out[0][0] \n conv5_block2_3_bn[0][0] \n__________________________________________________________________________________________________\nconv5_block2_out (Activation) (None, 7, 7, 2048) 0 conv5_block2_add[0][0] \n__________________________________________________________________________________________________\nconv5_block3_1_conv (Conv2D) (None, 7, 7, 512) 1049088 conv5_block2_out[0][0] \n__________________________________________________________________________________________________\nconv5_block3_1_bn (BatchNormali (None, 7, 7, 512) 2048 conv5_block3_1_conv[0][0] \n__________________________________________________________________________________________________\nconv5_block3_1_relu (Activation (None, 7, 7, 512) 0 conv5_block3_1_bn[0][0] \n__________________________________________________________________________________________________\nconv5_block3_2_conv (Conv2D) (None, 7, 7, 512) 2359808 conv5_block3_1_relu[0][0] \n__________________________________________________________________________________________________\nconv5_block3_2_bn (BatchNormali (None, 7, 7, 512) 2048 conv5_block3_2_conv[0][0] \n__________________________________________________________________________________________________\nconv5_block3_2_relu (Activation (None, 7, 7, 512) 0 conv5_block3_2_bn[0][0] \n__________________________________________________________________________________________________\nconv5_block3_3_conv (Conv2D) (None, 7, 7, 2048) 1050624 conv5_block3_2_relu[0][0] \n__________________________________________________________________________________________________\nconv5_block3_3_bn (BatchNormali (None, 7, 7, 2048) 8192 conv5_block3_3_conv[0][0] \n__________________________________________________________________________________________________\nconv5_block3_add (Add) (None, 7, 7, 2048) 0 conv5_block2_out[0][0] \n conv5_block3_3_bn[0][0] \n__________________________________________________________________________________________________\nconv5_block3_out (Activation) (None, 7, 7, 2048) 0 conv5_block3_add[0][0] \n__________________________________________________________________________________________________\nmax_pool (GlobalMaxPooling2D) (None, 2048) 0 conv5_block3_out[0][0] \n__________________________________________________________________________________________________\ndense_5 (Dense) (None, 512) 1049088 max_pool[0][0] \n__________________________________________________________________________________________________\nbatch_normalization_3 (BatchNor (None, 512) 2048 dense_5[0][0] \n__________________________________________________________________________________________________\ndense_6 (Dense) (None, 1) 513 batch_normalization_3[0][0] \n==================================================================================================\nTotal params: 24,639,361\nTrainable params: 24,585,217\nNon-trainable params: 54,144\n__________________________________________________________________________________________________\n" ], [ "# this is the augmentation configuration we will use for training\ntrain_datagen = ImageDataGenerator(\n rescale=1. / 255,\n shear_range=0.2,\n zoom_range=0.2,\n horizontal_flip=True,\n samplewise_center=True,\n samplewise_std_normalization=True,\n brightness_range=(0.1, 0.9))\n\n# this is the augmentation configuration we will use for testing:\n# only rescaling\ntest_datagen = ImageDataGenerator(rescale=1. / 255, samplewise_center=True,\n samplewise_std_normalization=True,)\n\ntrain_generator = train_datagen.flow_from_directory(\n train_data_dir,\n target_size=(img_width, img_height),\n batch_size=batch_size,\n class_mode='binary')\n\nvalidation_generator = test_datagen.flow_from_directory(\n validation_data_dir,\n target_size=(img_width, img_height),\n batch_size=batch_size,\n class_mode='binary')\n\n", "Found 193 images belonging to 2 classes.\nFound 49 images belonging to 2 classes.\n" ], [ "history = model.fit_generator(\n train_generator,\n steps_per_epoch=nb_train_samples // batch_size,\n epochs=epochs,\n validation_data=validation_generator,\n validation_steps=nb_validation_samples // batch_size)", "Epoch 1/50\n48/48 [==============================] - 40s 829ms/step - loss: 0.7163 - acc: 0.5988 - val_loss: 1.1285 - val_acc: 0.6042\nEpoch 2/50\n48/48 [==============================] - 29s 598ms/step - loss: 0.6657 - acc: 0.6718 - val_loss: 1.1587 - val_acc: 0.5778\nEpoch 3/50\n48/48 [==============================] - 29s 596ms/step - loss: 0.5581 - acc: 0.7291 - val_loss: 0.6846 - val_acc: 0.7333\nEpoch 4/50\n48/48 [==============================] - 28s 590ms/step - loss: 0.4439 - acc: 0.7659 - val_loss: 0.8754 - val_acc: 0.6889\nEpoch 5/50\n48/48 [==============================] - 28s 580ms/step - loss: 0.3797 - acc: 0.8023 - val_loss: 1.7015 - val_acc: 0.6444\nEpoch 6/50\n48/48 [==============================] - 28s 577ms/step - loss: 0.4168 - acc: 0.8073 - val_loss: 1.0582 - val_acc: 0.6889\nEpoch 7/50\n48/48 [==============================] - 27s 561ms/step - loss: 0.2777 - acc: 0.8854 - val_loss: 0.8164 - val_acc: 0.6889\nEpoch 8/50\n48/48 [==============================] - 27s 552ms/step - loss: 0.3054 - acc: 0.8749 - val_loss: 1.0977 - val_acc: 0.6222\nEpoch 9/50\n48/48 [==============================] - 26s 537ms/step - loss: 0.3164 - acc: 0.8593 - val_loss: 0.7639 - val_acc: 0.6667\nEpoch 10/50\n48/48 [==============================] - 26s 547ms/step - loss: 0.3216 - acc: 0.8492 - val_loss: 1.0854 - val_acc: 0.6000\nEpoch 11/50\n48/48 [==============================] - 25s 524ms/step - loss: 0.2744 - acc: 0.9062 - val_loss: 0.6132 - val_acc: 0.8000\nEpoch 12/50\n48/48 [==============================] - 27s 552ms/step - loss: 0.3335 - acc: 0.8437 - val_loss: 1.1778 - val_acc: 0.6667\nEpoch 13/50\n48/48 [==============================] - 26s 532ms/step - loss: 0.2455 - acc: 0.8649 - val_loss: 1.0283 - val_acc: 0.6222\nEpoch 14/50\n48/48 [==============================] - 27s 554ms/step - loss: 0.2898 - acc: 0.8906 - val_loss: 0.8862 - val_acc: 0.7083\nEpoch 15/50\n48/48 [==============================] - 29s 614ms/step - loss: 0.2921 - acc: 0.8701 - val_loss: 0.8799 - val_acc: 0.7111\nEpoch 16/50\n48/48 [==============================] - 26s 549ms/step - loss: 0.2659 - acc: 0.8958 - val_loss: 1.0852 - val_acc: 0.6000\nEpoch 17/50\n48/48 [==============================] - 27s 553ms/step - loss: 0.3294 - acc: 0.8958 - val_loss: 0.8373 - val_acc: 0.7333\nEpoch 18/50\n48/48 [==============================] - 26s 532ms/step - loss: 0.1963 - acc: 0.9323 - val_loss: 1.3287 - val_acc: 0.6444\nEpoch 19/50\n48/48 [==============================] - 21s 444ms/step - loss: 0.1924 - acc: 0.9166 - val_loss: 1.2395 - val_acc: 0.7111\nEpoch 20/50\n48/48 [==============================] - 19s 399ms/step - loss: 0.2159 - acc: 0.9323 - val_loss: 1.0421 - val_acc: 0.6889\nEpoch 21/50\n48/48 [==============================] - 19s 403ms/step - loss: 0.3113 - acc: 0.8440 - val_loss: 1.8189 - val_acc: 0.7111\nEpoch 22/50\n48/48 [==============================] - 19s 400ms/step - loss: 0.2216 - acc: 0.9167 - val_loss: 1.1871 - val_acc: 0.6667\nEpoch 23/50\n48/48 [==============================] - 18s 379ms/step - loss: 0.2921 - acc: 0.8752 - val_loss: 1.0717 - val_acc: 0.7111\nEpoch 24/50\n48/48 [==============================] - 19s 391ms/step - loss: 0.3003 - acc: 0.8909 - val_loss: 0.7622 - val_acc: 0.7111\nEpoch 25/50\n48/48 [==============================] - 19s 396ms/step - loss: 0.2722 - acc: 0.8854 - val_loss: 0.7565 - val_acc: 0.6889\nEpoch 26/50\n48/48 [==============================] - 18s 377ms/step - loss: 0.2215 - acc: 0.8802 - val_loss: 1.0120 - val_acc: 0.6222\nEpoch 27/50\n48/48 [==============================] - 20s 410ms/step - loss: 0.2242 - acc: 0.9062 - val_loss: 0.9107 - val_acc: 0.6875\nEpoch 28/50\n48/48 [==============================] - 21s 442ms/step - loss: 0.2708 - acc: 0.8645 - val_loss: 0.7850 - val_acc: 0.6444\nEpoch 29/50\n48/48 [==============================] - 19s 402ms/step - loss: 0.2649 - acc: 0.8857 - val_loss: 0.6502 - val_acc: 0.7556\nEpoch 30/50\n48/48 [==============================] - 19s 399ms/step - loss: 0.1987 - acc: 0.9323 - val_loss: 0.8587 - val_acc: 0.5556\nEpoch 31/50\n48/48 [==============================] - 20s 410ms/step - loss: 0.1854 - acc: 0.9169 - val_loss: 0.8794 - val_acc: 0.6222\nEpoch 32/50\n48/48 [==============================] - 18s 378ms/step - loss: 0.2531 - acc: 0.8909 - val_loss: 0.9173 - val_acc: 0.7556\nEpoch 33/50\n48/48 [==============================] - 21s 429ms/step - loss: 0.2019 - acc: 0.9479 - val_loss: 1.1515 - val_acc: 0.6000\nEpoch 34/50\n48/48 [==============================] - 20s 415ms/step - loss: 0.1721 - acc: 0.9378 - val_loss: 0.9395 - val_acc: 0.6000\nEpoch 35/50\n48/48 [==============================] - 21s 434ms/step - loss: 0.2007 - acc: 0.9218 - val_loss: 1.1680 - val_acc: 0.7111\nEpoch 36/50\n48/48 [==============================] - 25s 523ms/step - loss: 0.2500 - acc: 0.9062 - val_loss: 0.6188 - val_acc: 0.7333\nEpoch 37/50\n48/48 [==============================] - 24s 500ms/step - loss: 0.2523 - acc: 0.9118 - val_loss: 0.7521 - val_acc: 0.7333\nEpoch 38/50\n48/48 [==============================] - 23s 482ms/step - loss: 0.1867 - acc: 0.9218 - val_loss: 0.8352 - val_acc: 0.7556\nEpoch 39/50\n48/48 [==============================] - 24s 498ms/step - loss: 0.2207 - acc: 0.8857 - val_loss: 0.7915 - val_acc: 0.6667\nEpoch 40/50\n48/48 [==============================] - 24s 503ms/step - loss: 0.2219 - acc: 0.8961 - val_loss: 0.9396 - val_acc: 0.7292\nEpoch 41/50\n48/48 [==============================] - 29s 607ms/step - loss: 0.1833 - acc: 0.9271 - val_loss: 0.7488 - val_acc: 0.6889\nEpoch 42/50\n48/48 [==============================] - 25s 512ms/step - loss: 0.1286 - acc: 0.9323 - val_loss: 1.0086 - val_acc: 0.5778\nEpoch 43/50\n48/48 [==============================] - 26s 534ms/step - loss: 0.2698 - acc: 0.8906 - val_loss: 0.7913 - val_acc: 0.7111\nEpoch 44/50\n48/48 [==============================] - 25s 527ms/step - loss: 0.1924 - acc: 0.9065 - val_loss: 1.1655 - val_acc: 0.6667\nEpoch 45/50\n48/48 [==============================] - 28s 591ms/step - loss: 0.2192 - acc: 0.9166 - val_loss: 0.7385 - val_acc: 0.6889\nEpoch 46/50\n48/48 [==============================] - 20s 424ms/step - loss: 0.2735 - acc: 0.9114 - val_loss: 1.0243 - val_acc: 0.5333\nEpoch 47/50\n48/48 [==============================] - 20s 410ms/step - loss: 0.2474 - acc: 0.9271 - val_loss: 0.6580 - val_acc: 0.6889\nEpoch 48/50\n48/48 [==============================] - 23s 471ms/step - loss: 0.2075 - acc: 0.9118 - val_loss: 0.6266 - val_acc: 0.6667\nEpoch 49/50\n48/48 [==============================] - 23s 477ms/step - loss: 0.2088 - acc: 0.9427 - val_loss: 0.7290 - val_acc: 0.7111\nEpoch 50/50\n48/48 [==============================] - 21s 437ms/step - loss: 0.1955 - acc: 0.9427 - val_loss: 0.9983 - val_acc: 0.6222\n" ], [ "history.history.keys()", "_____no_output_____" ], [ "# summarize history for accuracy\nplt.plot(history.history['acc'])\nplt.plot(history.history['val_acc'])\nplt.title('model accuracy')\nplt.ylabel('accuracy')\nplt.xlabel('epoch')\nplt.legend(['train', 'test'], loc='upper left')\nplt.show()\n# summarize history for loss\nplt.plot(history.history['loss'])\nplt.plot(history.history['val_loss'])\nplt.title('model loss')\nplt.ylabel('loss')\nplt.xlabel('epoch')\nplt.legend(['train', 'test'], loc='upper left')\nplt.show()", "_____no_output_____" ], [ "from keras.preprocessing.image import ImageDataGenerator, array_to_img, img_to_array, load_img\n\ndatagen = ImageDataGenerator(\n rotation_range=40,\n width_shift_range=0.2,\n height_shift_range=0.2,\n shear_range=0.2,\n zoom_range=0.2,\n horizontal_flip=True,\n fill_mode='nearest',\n samplewise_center=True,\n samplewise_std_normalization=True,\n brightness_range=(0.1, 0.9))\n\nimg = load_img('../datasets/train/N/1_30_1_231.jpg') # this is a PIL image\nx = img_to_array(img) # this is a Numpy array with shape (3, 150, 150)\nx = x.reshape((1,) + x.shape) # this is a Numpy array with shape (1, 3, 150, 150)\n\n# the .flow() command below generates batches of randomly transformed images\n# and saves the results to the `preview/` directory\ni = 0\nfor batch in datagen.flow(x, batch_size=1,\n save_to_dir='preview', save_prefix='cat', save_format='jpeg'):\n i += 1\n if i > 20:\n break # otherwise the generator would loop indefinitely", "_____no_output_____" ], [ "x_train = []\nfor i in range(25):\n x_batch, y_batch = next(train_generator)\n x_train.append(x_batch)\n \nx_test = []\nfor i in range(2):\n x_batch, y_batch = next(validation_generator)\n x_test.append(x_batch)\n ", "_____no_output_____" ], [ "img = load_img('../dataset/test/N/1_30_2_382.JPG') # this is a PIL image\nplt.imshow(img)", "_____no_output_____" ], [ "imgNames = os.listdir('../dataset/test/N')\nfor img in imgNames: \n print(\"HERE\")\n imgname = '../dataset/test/N/' + img; \n myimg = load_img(imgname) # this is a PIL image\n plt.imshow(myimg)\n plt.show()", "_____no_output_____" ], [ "x_train = np.concatenate(x_train)\nx_train.shape\n\nx_test = np.concatenate(x_test)\nx_test.shape", "_____no_output_____" ], [ "import shap\nimport numpy as np\n\n# select a set of background examples to take an expectation over\nbackground = x_train[np.random.choice(x_train.shape[0], 10, replace=True)]\n\n# explain predictions of the model on four images\ne = shap.DeepExplainer(model, background)\n# ...or pass tensors directly\n# e = shap.DeepExplainer((model.layers[0].input, model.layers[-1].output), background)\nshap_values = e.shap_values(x_test[1:2])\n\n# plot the feature attributions\nshap.image_plot(shap_values, -x_test[1:2])", "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n" ], [ "x_test[1]", "_____no_output_____" ] ] ]

[ "markdown", "code" ]

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

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Leaflet cluster map of talk locations\n\nRun this from the _talks/ directory, which contains .md files of all your talks. This scrapes the location YAML field from each .md file, geolocates it with geopy/Nominatim, and uses the getorg library to output data, HTML, and Javascript for a standalone cluster map.", "_____no_output_____" ] ], [ [ "!pip install getorg --upgrade\nimport glob\nimport getorg\nfrom geopy import Nominatim", "Requirement already satisfied: getorg in c:\\users\\jialei shen\\appdata\\local\\programs\\python\\python37-32\\lib\\site-packages (0.3.1)\nRequirement already satisfied: geopy in c:\\users\\jialei shen\\appdata\\local\\programs\\python\\python37-32\\lib\\site-packages (from getorg) (2.2.0)\nRequirement already satisfied: pygithub in c:\\users\\jialei shen\\appdata\\local\\programs\\python\\python37-32\\lib\\site-packages (from getorg) (1.55)\nRequirement already satisfied: retrying in c:\\users\\jialei shen\\appdata\\local\\programs\\python\\python37-32\\lib\\site-packages (from getorg) (1.3.3)\nRequirement already satisfied: geographiclib<2,>=1.49 in c:\\users\\jialei shen\\appdata\\local\\programs\\python\\python37-32\\lib\\site-packages (from geopy->getorg) (1.52)\nRequirement already satisfied: deprecated in c:\\users\\jialei shen\\appdata\\local\\programs\\python\\python37-32\\lib\\site-packages (from pygithub->getorg) (1.2.13)\nRequirement already satisfied: requests>=2.14.0 in c:\\users\\jialei shen\\appdata\\local\\programs\\python\\python37-32\\lib\\site-packages (from pygithub->getorg) (2.22.0)\nRequirement already satisfied: pynacl>=1.4.0 in c:\\users\\jialei shen\\appdata\\local\\programs\\python\\python37-32\\lib\\site-packages (from pygithub->getorg) (1.5.0)\nRequirement already satisfied: pyjwt>=2.0 in c:\\users\\jialei shen\\appdata\\local\\programs\\python\\python37-32\\lib\\site-packages (from pygithub->getorg) (2.3.0)\nRequirement already satisfied: six>=1.7.0 in c:\\users\\jialei shen\\appdata\\local\\programs\\python\\python37-32\\lib\\site-packages (from retrying->getorg) (1.12.0)\nRequirement already satisfied: cffi>=1.4.1 in c:\\users\\jialei shen\\appdata\\local\\programs\\python\\python37-32\\lib\\site-packages (from pynacl>=1.4.0->pygithub->getorg) (1.12.3)\nRequirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in c:\\users\\jialei shen\\appdata\\local\\programs\\python\\python37-32\\lib\\site-packages (from requests>=2.14.0->pygithub->getorg) (1.25.3)\nRequirement already satisfied: certifi>=2017.4.17 in c:\\users\\jialei shen\\appdata\\local\\programs\\python\\python37-32\\lib\\site-packages (from requests>=2.14.0->pygithub->getorg) (2019.6.16)\nRequirement already satisfied: chardet<3.1.0,>=3.0.2 in c:\\users\\jialei shen\\appdata\\local\\programs\\python\\python37-32\\lib\\site-packages (from requests>=2.14.0->pygithub->getorg) (3.0.4)\nRequirement already satisfied: idna<2.9,>=2.5 in c:\\users\\jialei shen\\appdata\\local\\programs\\python\\python37-32\\lib\\site-packages (from requests>=2.14.0->pygithub->getorg) (2.8)\nRequirement already satisfied: wrapt<2,>=1.10 in c:\\users\\jialei shen\\appdata\\roaming\\python\\python37\\site-packages (from deprecated->pygithub->getorg) (1.12.1)\nRequirement already satisfied: pycparser in c:\\users\\jialei shen\\appdata\\local\\programs\\python\\python37-32\\lib\\site-packages (from cffi>=1.4.1->pynacl>=1.4.0->pygithub->getorg) (2.19)\nIywidgets and ipyleaflet support disabled. You must be in a Jupyter notebook to use this feature.\nError raised:\nNo module named 'ipyleaflet'\nCheck that you have enabled ipyleaflet in Jupyter with:\n jupyter nbextension enable --py ipyleaflet\n" ], [ "g = glob.glob(\"*.md\")", "_____no_output_____" ], [ "geocoder = Nominatim()\nlocation_dict = {}\nlocation = \"\"\npermalink = \"\"\ntitle = \"\"", "_____no_output_____" ], [ "\nfor file in g:\n with open(file, 'r') as f:\n lines = f.read()\n if lines.find('location: \"') > 1:\n loc_start = lines.find('location: \"') + 11\n lines_trim = lines[loc_start:]\n loc_end = lines_trim.find('\"')\n location = lines_trim[:loc_end]\n \n \n location_dict[location] = geocoder.geocode(location)\n print(location, \"\\n\", location_dict[location])\n", "_____no_output_____" ], [ "m = getorg.orgmap.create_map_obj()\ngetorg.orgmap.output_html_cluster_map(location_dict, folder_name=\"../talkmap\", hashed_usernames=False)", "_____no_output_____" ] ] ]

[ "markdown", "code" ]

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

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ [ [ "import sklearn\nprint(sklearn)\nfrom sklearn.tree import LinearDecisionTreeRegressor as ldtr\nfrom sklearn.linear_model import LinearRegression\nfrom sklearn.linear_model import Lasso\nfrom sklearn.svm import LinearSVR\nfrom sklearn.svm import SVR\nfrom sklearn.tree import DecisionTreeRegressor\nfrom sklearn.linear_model import SGDRegressor\nfrom sklearn.ensemble import RandomForestRegressor\nfrom sklearn.feature_selection import SelectKBest, f_regression\nimport seaborn as sns\nimport numpy as np\nimport pandas as pd\nfrom matplotlib.ticker import FormatStrFormatter\nimport matplotlib.pyplot as plt\nfrom sklearn.metrics import mean_squared_error\nfrom sklearn.metrics import mean_absolute_error\nfrom sklearn import datasets\nfrom sklearn.pipeline import make_pipeline\nfrom sklearn.preprocessing import StandardScaler\nfrom sklearn.datasets import make_regression\nfrom sklearn.model_selection import train_test_split\nfrom sklearn.ensemble import GradientBoostingRegressor\nplt.style.use('seaborn-whitegrid')", "<module 'sklearn' from 'B:\\\\Research\\\\Decision Tree Regressor Combined with Linear Regressor\\\\code\\\\sklearn\\\\__init__.py'>\n" ], [ "from sklearn.preprocessing import StandardScaler", "_____no_output_____" ], [ "scaler = StandardScaler()", "_____no_output_____" ], [ "df= pd.read_excel('./Concrete/Concrete_Data.xls')\ndf.head()", "_____no_output_____" ], [ "df.corr()", "_____no_output_____" ], [ "df.columns", "_____no_output_____" ], [ "X = np.array(df.drop(['Concrete compressive strength(MPa, megapascals) '], axis= 1))\n# X = X[:, [0, 2, 6]]\ny = np.array(df['Concrete compressive strength(MPa, megapascals) '])\n\n\nX= scaler.fit_transform(X)\n\n# indices = X[:, 2] > -1\n# y= y[indices]\n# X= X[indices]\n\n\nX_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)", "_____no_output_____" ], [ "# for i in range(X_train.shape[1]):\n# plt.scatter(X[:, i], y, color='black')\n# plt.show()", "_____no_output_____" ], [ "X", "_____no_output_____" ], [ "print(type(X_train))\nprint(X_train.shape)\nprint(X_test.shape)\nprint(y_train.shape)\nprint(y_test.shape)", "<class 'numpy.ndarray'>\n(824, 8)\n(206, 8)\n(824,)\n(206,)\n" ], [ "reg1 = ldtr(max_depth= 2)\nreg2 = LinearRegression(n_jobs = -1)\nreg3 = LinearSVR()\nreg4 = SVR(kernel = 'rbf')\nreg5 = DecisionTreeRegressor(max_depth = 2)\nreg6 = Lasso()\nreg7 = RandomForestRegressor(max_depth = 2,n_jobs = -1)", "_____no_output_____" ], [ "reg1.fit(X_train, y_train)\nreg2.fit(X_train, y_train)\nreg3.fit(X_train, y_train)\nreg4.fit(X_train, y_train)\nreg5.fit(X_train, y_train)\nreg6.fit(X_train, y_train)\nreg7.fit(X_train, y_train)", "_____no_output_____" ], [ "reg2.coef_", "_____no_output_____" ], [ "y_pred1 = reg1.predict(X_test)\ny_pred2 = reg2.predict(X_test)\ny_pred3 = reg3.predict(X_test)\ny_pred4 = reg4.predict(X_test)\ny_pred5 = reg5.predict(X_test)\ny_pred6 = reg6.predict(X_test)\ny_pred7 = reg7.predict(X_test)", "_____no_output_____" ], [ "print('Mean squared error: %.2f'% np.sqrt(mean_squared_error(y_test, y_pred1)))\nprint('Mean squared error: %.2f'% mean_squared_error(y_test, y_pred2))\nprint('Mean squared error: %.2f'% mean_squared_error(y_test, y_pred3))\nprint('Mean squared error: %.2f'% mean_squared_error(y_test, y_pred4))\nprint('Mean squared error: %.2f'% mean_squared_error(y_test, y_pred5))\nprint('Mean squared error: %.2f'% mean_squared_error(y_test, y_pred6))\nprint('Mean squared error: %.2f'% mean_squared_error(y_test, y_pred7))", "Mean squared error: 6.06\nMean squared error: 116.62\nMean squared error: 118.69\nMean squared error: 103.85\nMean squared error: 147.99\nMean squared error: 135.02\nMean squared error: 133.70\n" ], [ "def regression_model(model):\n regressor = model\n regressor.fit(X_train,y_train)\n y_pred = regressor.predict(X_test)\n rmse = np.sqrt(mean_squared_error(y_test, y_pred))\n mae = mean_absolute_error(y_test, y_pred)\n return rmse,mae\n\n\nmodel_performance = pd.DataFrame(columns = [\"Model\",\"RMSE\",\"MAE\"])\nmodel_to_evaluate = [ldtr(max_depth= 1),\n LinearRegression(n_jobs = -1),\n SVR(kernel = 'rbf'),\n DecisionTreeRegressor(max_depth = 1, min_samples_leaf= 3, min_samples_split= 6),\n Lasso(),\n RandomForestRegressor(max_depth = 1,n_jobs = -1)]\n\nfor model in model_to_evaluate:\n rmse,mae = regression_model(model)\n model_performance = model_performance.append({\"Model\" : model,\"RMSE\" : rmse,\"MAE\": mae},ignore_index = True)", "_____no_output_____" ], [ "model_performance", "_____no_output_____" ], [ "labels = ['LDTR','LR','SVR','DT','Lasso','RFR']\n\nx = np.arange(len(labels))\nwidth = 0.3\n\nfig, ax = plt.subplots()\nrects1 = ax.bar(x - width/2, model_performance['RMSE'], width, label='RMSE')\nrects2 = ax.bar(x + width/2, model_performance['MAE'], width, label='MAE')\n\nax.set_ylabel('Error',fontweight = 'bold')\nax.set_title('RMSE and MAE for Concrete Dataset')\nax.set_xticks(x)\nax.set_xticklabels(labels,fontweight = 'bold')\n# ax.set_yticklabels([0.0,2.0,4.0,6.0,8.0,10.0,12.0,14.0,16.0])\nax.yaxis.set_major_formatter(FormatStrFormatter('%.1f'))\nax.legend()\n\n\nfig.tight_layout()\nplt.savefig('B:/concrete.png')\nplt.show()", "_____no_output_____" ], [ "for i in range(X_train.shape[1]):\n plt.scatter(X_train[:, i], y_train, color='black')\n plt.scatter(X_train[:, i], reg1.predict(X_train), color='blue')\n plt.scatter(X_train[:, i], reg6.predict(X_train), color='green')\n plt.show()", "_____no_output_____" ], [ "for i in range(X_train.shape[1]):\n plt.scatter(X_test[:, i], y_test, color='black')\n plt.scatter(X_test[:, i], reg1.predict(X_test), color='blue')\n plt.scatter(X_train[:, i], reg2.predict(X_train), color='green')\n plt.show()", "_____no_output_____" ], [ "np.sum(np.array([1, 2]) * np.array([1, 2]))", "_____no_output_____" ], [ "%%time\nreg1.fit(X_train, y_train)", "_____no_output_____" ] ] ]

[ "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" ]

[ [ [ "import numpy as np\nimport pandas as pd\nfrom pathlib import Path\n%matplotlib inline", "_____no_output_____" ] ], [ [ "# Regression Analysis: Seasonal Effects with Sklearn Linear Regression\nIn this notebook, you will build a SKLearn linear regression model to predict Yen futures (\"settle\") returns with *lagged* Yen futures returns. ", "_____no_output_____" ] ], [ [ "# Futures contract on the Yen-dollar exchange rate:\n# This is the continuous chain of the futures contracts that are 1 month to expiration\nyen_futures = pd.read_csv(\n Path(\"./data/yen.csv\"), index_col=\"Date\", infer_datetime_format=True, parse_dates=True\n)\nyen_futures.head()", "_____no_output_____" ], [ "# Trim the dataset to begin on January 1st, 1990\nyen_futures = yen_futures.loc[\"1990-01-01\":, :]\nyen_futures.head()", "_____no_output_____" ] ], [ [ "# Data Preparation", "_____no_output_____" ], [ "### Returns", "_____no_output_____" ] ], [ [ "# Create a series using \"Settle\" price percentage returns, drop any nan\"s, and check the results:\n# (Make sure to multiply the pct_change() results by 100)\n# In this case, you may have to replace inf, -inf values with np.nan\"s\nyen_futures['Return'] = (yen_futures['Settle'].pct_change() *100)\nyen_futures = yen_futures.replace(-np.inf, np.nan).dropna()\nyen_futures.tail()", "_____no_output_____" ] ], [ [ "### Lagged Returns ", "_____no_output_____" ] ], [ [ "# Create a lagged return using the shift function\nyen_futures['Lagged_Return'] = yen_futures['Return'].shift()\nyen_futures = yen_futures.dropna()\nyen_futures.tail()", "_____no_output_____" ] ], [ [ "### Train Test Split", "_____no_output_____" ] ], [ [ "# Create a train/test split for the data using 2018-2019 for testing and the rest for training\ntrain = yen_futures[:'2017']\ntest = yen_futures['2018':]", "_____no_output_____" ], [ "# Create four dataframes:\n# X_train (training set using just the independent variables), X_test (test set of just the independent variables)\n# Y_train (training set using just the \"y\" variable, i.e., \"Futures Return\"), Y_test (test set of just the \"y\" variable):\nX_train = train['Lagged_Return'].to_frame()\nX_test = test['Lagged_Return'].to_frame()\ny_train = train['Return']\ny_test = test['Return']\n", "_____no_output_____" ], [ "X_train\n", "_____no_output_____" ] ], [ [ "# Linear Regression Model", "_____no_output_____" ] ], [ [ "# Create a Linear Regression model and fit it to the training data\nfrom sklearn.linear_model import LinearRegression\n\n# Fit a SKLearn linear regression using just the training set (X_train, Y_train):\nmodel = LinearRegression(copy_X=True, fit_intercept=True, n_jobs=None, normalize=False)\nmodel.fit(X_train, y_train)\n", "_____no_output_____" ] ], [ [ "# Make predictions using the Testing Data\n\nNote: We want to evaluate the model using data that it has never seen before, in this case: X_test.", "_____no_output_____" ] ], [ [ "# Make a prediction of \"y\" values using just the test dataset\npredictions = model.predict(X_test)\n", "_____no_output_____" ], [ "# Assemble actual y data (Y_test) with predicted y data (from just above) into two columns in a dataframe:\nresults = y_test.to_frame()\nresults['Predicted Return'] = predictions\nresults.head()\n", "_____no_output_____" ], [ "# Plot the first 20 predictions vs the true values\nresults[:20].plot(title='Return vs Predicted Return', subplots=True, figsize=(12,8))\n", "_____no_output_____" ] ], [ [ "# Out-of-Sample Performance\n\nEvaluate the model using \"out-of-sample\" data (X_test and y_test)", "_____no_output_____" ] ], [ [ "from sklearn.metrics import mean_squared_error, r2_score\n# Calculate the mean_squared_error (MSE) on actual versus predicted test \"y\"\nmse = mean_squared_error(results['Return'], results['Predicted Return'])\n\n# Using that mean-squared-error, calculate the root-mean-squared error (RMSE):\nrmse = np.sqrt(mse)\nprint(f'Out-of-Sample Root Mean Squared Error (RMSE): {rmse}')\n", "Out-of-Sample Root Mean Squared Error (RMSE): 0.4154832784856737\n" ] ], [ [ "# In-Sample Performance\n\nEvaluate the model using in-sample data (X_train and y_train)", "_____no_output_____" ] ], [ [ "# Construct a dataframe using just the \"y\" training data:\ndf_in_sample_results = y_train.to_frame()\n\n# Add a column of \"in-sample\" predictions to that dataframe: \ndf_in_sample_results['In-Sample'] = model.predict(X_train)\n\n# Calculate in-sample mean_squared_error (for comparison to out-of-sample)\nmse = mean_squared_error(df_in_sample_results['Return'], df_in_sample_results['In-Sample'])\n\n# Calculate in-sample root mean_squared_error (for comparison to out-of-sample)\nrmse = np.sqrt(mse)\n\n# Output root mean squared error\nprint(f'In-Sample Root Mean Squared Error (RMSE): {rmse}')\n", "In-Sample Root Mean Squared Error (RMSE): 0.5963660785073426\n" ] ], [ [ "# Conclusions", "_____no_output_____" ], [ "Based on the analysis and outcomes, the following information is known.\n- The out-of-sample data shows a root mean squared error (RMSE) of 0.4154832784856737.\n- The in-sample data shows a root mean squared error (RMSE) of 0.5963660785073426.\n\nBased on a comparison of these two values, the results show the model performs better when trained on data it has not been trained on.\n\nIn summary, further analysis and testing is needed such as change the sample time frame to provide when training the model.\n", "_____no_output_____" ] ] ]

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

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

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import pandas as pd\ndataframe = pd.read_csv(r\"C:\\Users\\Abhishek\\Desktop\\Computer Science Texas Tech University\\Fall 2018\\CS 5341 Pattern Recognition\\Healthcare Project\\train_2v.csv\")\n", "_____no_output_____" ], [ "print(dataframe.head(5))", " id gender age hypertension heart_disease ever_married \\\n0 30669 Male 3.0 0 0 No \n1 30468 Male 58.0 1 0 Yes \n2 16523 Female 8.0 0 0 No \n3 56543 Female 70.0 0 0 Yes \n4 46136 Male 14.0 0 0 No \n\n work_type Residence_type avg_glucose_level bmi smoking_status \\\n0 children Rural 95.12 18.0 NaN \n1 Private Urban 87.96 39.2 never smoked \n2 Private Urban 110.89 17.6 NaN \n3 Private Rural 69.04 35.9 formerly smoked \n4 Never_worked Rural 161.28 19.1 NaN \n\n stroke \n0 0 \n1 0 \n2 0 \n3 0 \n4 0 \n" ], [ "# take a look at the outcome variable stroke\nprint(dataframe['stroke'].value_counts())", "0 42617\n1 783\nName: stroke, dtype: int64\n" ], [ "print(dataframe.isnull().sum().sort_values(ascending=False).head())", "smoking_status 13292\nbmi 1462\nstroke 0\navg_glucose_level 0\nResidence_type 0\ndtype: int64\n" ], [ "# removing nans from bmi using mean or median\nprint(dataframe['bmi'].mean())\nprint(dataframe['bmi'].median())", "28.605038390004545\n27.7\n" ], [ "# decided to replace with mean\ndataframe['bmi'].fillna(dataframe['bmi'].mean(), inplace=True)", "_____no_output_____" ], [ "print(dataframe.isnull().sum().sort_values(ascending=False).head())", "smoking_status 13292\nstroke 0\nbmi 0\navg_glucose_level 0\nResidence_type 0\ndtype: int64\n" ], [ "dataframe = dataframe.dropna(subset=['smoking_status']) ", "_____no_output_____" ], [ "print(dataframe.isnull().sum().sort_values(ascending=False).head())", "stroke 0\nsmoking_status 0\nbmi 0\navg_glucose_level 0\nResidence_type 0\ndtype: int64\n" ], [ "dataframe_clean = dataframe\nprint (dataframe_clean.head(5))", " id gender age hypertension heart_disease ever_married \\\n1 30468 Male 58.0 1 0 Yes \n3 56543 Female 70.0 0 0 Yes \n6 52800 Female 52.0 0 0 Yes \n7 41413 Female 75.0 0 1 Yes \n8 15266 Female 32.0 0 0 Yes \n\n work_type Residence_type avg_glucose_level bmi smoking_status \\\n1 Private Urban 87.96 39.2 never smoked \n3 Private Rural 69.04 35.9 formerly smoked \n6 Private Urban 77.59 17.7 formerly smoked \n7 Self-employed Rural 243.53 27.0 never smoked \n8 Private Rural 77.67 32.3 smokes \n\n stroke \n1 0 \n3 0 \n6 0 \n7 0 \n8 0 \n" ], [ "def gendervalue(a):\n if a == 'Male':\n return 1\n if a == 'Female':\n return 0\n if a == 'Other':\n return 3\n \ndataframe_clean['gender'] = dataframe_clean['gender'].apply(gendervalue)\n \n\ndef work_type_class(a):\n if a == 'Private':\n return 0\n if a == 'Self-employed':\n return 1\n if a == 'children':\n return 2\n if a == 'Govt_job':\n return 3\n if a == 'Never_worked':\n return 4\n \ndataframe_clean['work_type'] = dataframe_clean['work_type'].apply(work_type_class)\n\n\ndef ever_married_value(a):\n if a == 'Yes':\n return 1\n if a == 'No':\n return 0\n \ndataframe_clean['ever_married'] = dataframe_clean['ever_married'].apply(ever_married_value)", "_____no_output_____" ], [ "def Residence_type_value(a):\n if a == 'Urban':\n return 1\n if a == 'Rural':\n return 0\n \ndataframe_clean['Residence_type'] = dataframe_clean['Residence_type'].apply(Residence_type_value)\n\ndef smoking_status_class(a):\n if a == 'never smoked':\n return 0\n if a == 'formerly smoked':\n return 1\n if a == 'smokes':\n return 2\n \ndataframe_clean['smoking_status'] = dataframe_clean['smoking_status'].apply(smoking_status_class)\n\nprint(dataframe_clean.head(5))", " id gender age hypertension heart_disease ever_married work_type \\\n1 30468 1 58.0 1 0 1 0 \n3 56543 0 70.0 0 0 1 0 \n6 52800 0 52.0 0 0 1 0 \n7 41413 0 75.0 0 1 1 1 \n8 15266 0 32.0 0 0 1 0 \n\n Residence_type avg_glucose_level bmi smoking_status stroke \n1 1 87.96 39.2 0 0 \n3 0 69.04 35.9 1 0 \n6 1 77.59 17.7 1 0 \n7 0 243.53 27.0 0 0 \n8 0 77.67 32.3 2 0 \n" ], [ "import numpy as np\nfrom pandas import Series, DataFrame\n\nfrom pandas.tools.plotting import scatter_matrix\n\nimport matplotlib.pyplot as plt\nfrom pylab import rcParams\nimport seaborn as sb", "_____no_output_____" ], [ "%matplotlib inline\nrcParams['figure.figsize']=10,8\nsb.set_style('whitegrid')", "_____no_output_____" ], [ "dataframe_clean.columns = ['id','gender','age','hypertension','heart_disease','ever_married','work_type','Residence_type','avg_glucose_level', 'bmi','smoking_status','stroke']\ndataframe_clean.index = dataframe_clean.id\nbmi = dataframe_clean['bmi']\nplt.xlabel('bmi')\nplt.ylabel('frequency')\nbmi.plot(kind ='hist')", "_____no_output_____" ], [ "avg_glucose_level = dataframe_clean['avg_glucose_level']\nplt.xlabel('bmi')\nplt.ylabel('frequency')\navg_glucose_level.plot(kind ='hist')", "_____no_output_____" ], [ "dataframe_clean.plot(kind='scatter', x='id', y='bmi', c=['darkgray'], s=1)", "_____no_output_____" ], [ "dataframe.plot(kind='scatter', x='id', y='avg_glucose_level', c=['darkgray'], s=1)", "_____no_output_____" ], [ "sb.regplot(x='id', y='bmi', data=dataframe_clean, scatter=True, scatter_kws={'s':2}) ", "C:\\Users\\Abhishek\\Anaconda3\\lib\\site-packages\\scipy\\stats\\stats.py:1713: FutureWarning: Using a non-tuple sequence for multidimensional indexing is deprecated; use `arr[tuple(seq)]` instead of `arr[seq]`. In the future this will be interpreted as an array index, `arr[np.array(seq)]`, which will result either in an error or a different result.\n return np.add.reduce(sorted[indexer] * weights, axis=axis) / sumval\n" ], [ "sb.regplot(x='id', y='avg_glucose_level', data=dataframe_clean, scatter=True, scatter_kws={'s':2}) ", "C:\\Users\\Abhishek\\Anaconda3\\lib\\site-packages\\scipy\\stats\\stats.py:1713: FutureWarning: Using a non-tuple sequence for multidimensional indexing is deprecated; use `arr[tuple(seq)]` instead of `arr[seq]`. In the future this will be interpreted as an array index, `arr[np.array(seq)]`, which will result either in an error or a different result.\n return np.add.reduce(sorted[indexer] * weights, axis=axis) / sumval\n" ], [ "from pandas import Series, DataFrame\n\nfrom pandas.tools.plotting import scatter_matrix", "_____no_output_____" ], [ "%matplotlib inline\nrcParams['figure.figsize']=14,7\nsb.set_style('whitegrid')", "_____no_output_____" ], [ "dataframe_new = dataframe.drop('id', 1)\nprint(dataframe_new.head(5))", " gender age hypertension heart_disease ever_married work_type \\\nid \n30468 1 58.0 1 0 1 0 \n56543 0 70.0 0 0 1 0 \n52800 0 52.0 0 0 1 0 \n41413 0 75.0 0 1 1 1 \n15266 0 32.0 0 0 1 0 \n\n Residence_type avg_glucose_level bmi smoking_status stroke \nid \n30468 1 87.96 39.2 0 0 \n56543 0 69.04 35.9 1 0 \n52800 1 77.59 17.7 1 0 \n41413 0 243.53 27.0 0 0 \n15266 0 77.67 32.3 2 0 \n" ], [ "dataframe_clean_new = dataframe_clean.drop('id', 1)\nprint(dataframe_clean_new.head(5))", " gender age hypertension heart_disease ever_married work_type \\\nid \n30468 1 58.0 1 0 1 0 \n56543 0 70.0 0 0 1 0 \n52800 0 52.0 0 0 1 0 \n41413 0 75.0 0 1 1 1 \n15266 0 32.0 0 0 1 0 \n\n Residence_type avg_glucose_level bmi smoking_status stroke \nid \n30468 1 87.96 39.2 0 0 \n56543 0 69.04 35.9 1 0 \n52800 1 77.59 17.7 1 0 \n41413 0 243.53 27.0 0 0 \n15266 0 77.67 32.3 2 0 \n" ], [ "green_diamond = dict(markerfacecolor='g', marker='D')\ndataframe_clean_new.boxplot(flierprops=green_diamond)\nplt.plot()", "_____no_output_____" ], [ "dataframe_clean.columns = ['id','gender','age','hypertension','heart_disease','ever_married','work_type','Residence_type','avg_glucose_level', 'bmi','smoking_status','stroke']\nX = dataframe_clean.iloc[:,0:10].values\nY = dataframe_clean.iloc[:,10].values\n\ndataframe_clean[:5]\nprint (X)", "[[3.0468e+04 1.0000e+00 5.8000e+01 ... 1.0000e+00 8.7960e+01 3.9200e+01]\n [5.6543e+04 0.0000e+00 7.0000e+01 ... 0.0000e+00 6.9040e+01 3.5900e+01]\n [5.2800e+04 0.0000e+00 5.2000e+01 ... 1.0000e+00 7.7590e+01 1.7700e+01]\n ...\n [2.8375e+04 0.0000e+00 8.2000e+01 ... 1.0000e+00 9.1940e+01 2.8900e+01]\n [2.7973e+04 1.0000e+00 4.0000e+01 ... 1.0000e+00 9.9160e+01 3.3200e+01]\n [3.6271e+04 0.0000e+00 8.2000e+01 ... 1.0000e+00 7.9480e+01 2.0600e+01]]\n" ], [ "pd.options.display.float_format = '{:,.1f}'.format\n\nX_df = pd.DataFrame(X)\n\nprint (X_df.describe())", " 0 1 2 3 4 5 6 7 \\\ncount 30,108.0 30,108.0 30,108.0 30,108.0 30,108.0 30,108.0 30,108.0 30,108.0 \nmean 36,596.9 0.4 47.9 0.1 0.1 0.7 0.7 0.5 \nstd 21,119.3 0.5 18.8 0.3 0.2 0.4 1.1 0.5 \nmin 1.0 0.0 10.0 0.0 0.0 0.0 0.0 0.0 \n25% 18,254.2 0.0 33.0 0.0 0.0 0.0 0.0 0.0 \n50% 36,775.5 0.0 48.0 0.0 0.0 1.0 0.0 1.0 \n75% 54,861.8 1.0 62.0 0.0 0.0 1.0 1.0 1.0 \nmax 72,943.0 3.0 82.0 1.0 1.0 1.0 4.0 1.0 \n\n 8 9 \ncount 30,108.0 30,108.0 \nmean 107.2 30.0 \nstd 46.0 7.1 \nmin 55.0 10.1 \n25% 77.8 25.1 \n50% 92.4 28.7 \n75% 114.5 33.6 \nmax 291.1 92.0 \n" ], [ "x = dataframe.drop('stroke', 1)\nprint(x.head(5))", " id gender age hypertension heart_disease ever_married \\\nid \n30468 30468 1 58.0 1 0 1 \n56543 56543 0 70.0 0 0 1 \n52800 52800 0 52.0 0 0 1 \n41413 41413 0 75.0 0 1 1 \n15266 15266 0 32.0 0 0 1 \n\n work_type Residence_type avg_glucose_level bmi smoking_status \nid \n30468 0 1 88.0 39.2 0 \n56543 0 0 69.0 35.9 1 \n52800 0 1 77.6 17.7 1 \n41413 1 0 243.5 27.0 0 \n15266 0 0 77.7 32.3 2 \n" ], [ "y = dataframe['stroke']\nprint(y.head(5))", "id\n30468 0\n56543 0\n52800 0\n41413 0\n15266 0\nName: stroke, dtype: int64\n" ], [ "import numpy as np\nimport pandas as pd\nimport matplotlib.pyplot as plt\ndef plot_histogram_dv(x, y):\n plt.hist(list(x[y == 0]), alpha=0.5, label='Stroke=0')\n plt.hist(list(x[y == 1]), alpha=0.5, label='Stroke=1')\n plt.title(\"Histogram of '{var_name}' by Stroke Category\".format(var_name=x.name))\n plt.xlabel(\"Value\")\n plt.ylabel(\"Frequency\")\n plt.legend(loc='upper right')\n plt.show()\n\n\nplot_histogram_dv(x['age'], y)\nplot_histogram_dv(x['bmi'], y)\nplot_histogram_dv(x['hypertension'], y)\nplot_histogram_dv(x['avg_glucose_level'], y)", "_____no_output_____" ] ] ]

[ "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" ]

[ [ [ "# DAT210x - Programming with Python for DS", "_____no_output_____" ], [ "## Module5- Lab6", "_____no_output_____" ] ], [ [ "import random, math\nimport pandas as pd\nimport numpy as np\nimport scipy.io\nfrom sklearn.model_selection import train_test_split\nfrom sklearn.decomposition import PCA\nfrom sklearn import manifold\nfrom sklearn.neighbors import KNeighborsClassifier\n\nfrom mpl_toolkits.mplot3d import Axes3D\nimport matplotlib\nimport matplotlib.pyplot as plt\n\nmatplotlib.style.use('ggplot') # Look Pretty\n\n\n# Leave this alone until indicated:\nTest_PCA = False", "_____no_output_____" ] ], [ [ "### A Convenience Function", "_____no_output_____" ], [ "This method is for your visualization convenience only. You aren't expected to know how to put this together yourself, although you should be able to follow the code by now:", "_____no_output_____" ] ], [ [ "def Plot2DBoundary(DTrain, LTrain, DTest, LTest):\n # The dots are training samples (img not drawn), and the pics are testing samples (images drawn)\n # Play around with the K values. This is very controlled dataset so it should be able to get perfect classification on testing entries\n # Play with the K for isomap, play with the K for neighbors. \n\n fig = plt.figure()\n ax = fig.add_subplot(111)\n ax.set_title('Transformed Boundary, Image Space -> 2D')\n\n padding = 0.1 # Zoom out\n resolution = 1 # Don't get too detailed; smaller values (finer rez) will take longer to compute\n colors = ['blue','green','orange','red']\n\n\n # ------\n\n # Calculate the boundaries of the mesh grid. The mesh grid is\n # a standard grid (think graph paper), where each point will be\n # sent to the classifier (KNeighbors) to predict what class it\n # belongs to. This is why KNeighbors has to be trained against\n # 2D data, so we can produce this countour. Once we have the \n # label for each point on the grid, we can color it appropriately\n # and plot it.\n x_min, x_max = DTrain[:, 0].min(), DTrain[:, 0].max()\n y_min, y_max = DTrain[:, 1].min(), DTrain[:, 1].max()\n x_range = x_max - x_min\n y_range = y_max - y_min\n x_min -= x_range * padding\n y_min -= y_range * padding\n x_max += x_range * padding\n y_max += y_range * padding\n\n # Using the boundaries, actually make the 2D Grid Matrix:\n xx, yy = np.meshgrid(np.arange(x_min, x_max, resolution),\n np.arange(y_min, y_max, resolution))\n\n # What class does the classifier say about each spot on the chart?\n # The values stored in the matrix are the predictions of the model\n # at said location:\n Z = model.predict(np.c_[xx.ravel(), yy.ravel()])\n Z = Z.reshape(xx.shape)\n\n # Plot the mesh grid as a filled contour plot:\n plt.contourf(xx, yy, Z, cmap=plt.cm.terrain, z=-100)\n\n\n # ------\n\n # When plotting the testing images, used to validate if the algorithm\n # is functioning correctly, size them as 5% of the overall chart size\n x_size = x_range * 0.05\n y_size = y_range * 0.05\n\n # First, plot the images in your TEST dataset\n img_num = 0\n for index in LTest.index:\n # DTest is a regular NDArray, so you'll iterate over that 1 at a time.\n x0, y0 = DTest[img_num,0]-x_size/2., DTest[img_num,1]-y_size/2.\n x1, y1 = DTest[img_num,0]+x_size/2., DTest[img_num,1]+y_size/2.\n\n # DTest = our images isomap-transformed into 2D. But we still want\n # to plot the original image, so we look to the original, untouched\n # dataset (at index) to get the pixels:\n img = df.iloc[index,:].reshape(num_pixels, num_pixels)\n ax.imshow(img,\n aspect='auto',\n cmap=plt.cm.gray,\n interpolation='nearest',\n zorder=100000,\n extent=(x0, x1, y0, y1),\n alpha=0.8)\n img_num += 1\n\n\n # Plot your TRAINING points as well... as points rather than as images\n for label in range(len(np.unique(LTrain))):\n indices = np.where(LTrain == label)\n ax.scatter(DTrain[indices, 0], DTrain[indices, 1], c=colors[label], alpha=0.8, marker='o')\n\n # Plot\n plt.show() ", "_____no_output_____" ] ], [ [ "### The Assignment", "_____no_output_____" ], [ "Use the same code from Module4/assignment4.ipynb to load up the `face_data.mat` file into a dataframe called `df`. Be sure to calculate the `num_pixels` value, and to rotate the images to being right-side-up instead of sideways. This was demonstrated in the [Lab Assignment 4](https://github.com/authman/DAT210x/blob/master/Module4/assignment4.ipynb) code.", "_____no_output_____" ] ], [ [ "# .. your code here ..\nmat = scipy.io.loadmat('Datasets/face_data.mat')\ndf = pd.DataFrame(mat['images']).T\nnum_images, num_pixels = df.shape\nnum_pixels = int(math.sqrt(num_pixels))\n\n# Rotate the pictures, so we don't have to crane our necks:\nfor i in range(num_images):\n df.loc[i,:] = df.loc[i,:].reshape(num_pixels, num_pixels).T.reshape(-1)", "_____no_output_____" ] ], [ [ "Load up your face_labels dataset. It only has a single column, and you're only interested in that single column. You will have to slice the column out so that you have access to it as a \"Series\" rather than as a \"Dataframe\". This was discussed in the the \"Slicin'\" lecture of the \"Manipulating Data\" reading on the course website. Use an appropriate indexer to take care of that. Be sure to print out the labels and compare what you see to the raw `face_labels.csv` so you know you loaded it correctly.", "_____no_output_____" ] ], [ [ "# .. your code here ..\nface_labels = pd.read_csv('Datasets/face_labels.csv',header=None)\nlabel = face_labels.iloc[:, 0]\nlen(label)", "_____no_output_____" ], [ "df.shape\ndf.head()", "_____no_output_____" ], [ "y.head", "_____no_output_____" ] ], [ [ "Do `train_test_split`. Use the same code as on the EdX platform in the reading material, but set the random_state=7 for reproducibility, and the test_size to 0.15 (150%). Your labels are actually passed in as a series (instead of as an NDArray) so that you can access their underlying indices later on. This is necessary so you can find your samples in the original dataframe. The convenience methods we've written for you that handle drawing expect this, so that they can plot your testing data as images rather than as points:", "_____no_output_____" ] ], [ [ "# .. your code here ..\nx_train, x_test, y_train, y_test = train_test_split(df, label, test_size=0.15, random_state=7)", "_____no_output_____" ] ], [ [ "### Dimensionality Reduction", "_____no_output_____" ] ], [ [ "if Test_PCA:\n # INFO: PCA is used *before* KNeighbors to simplify your high dimensionality\n # image samples down to just 2 principal components! A lot of information\n # (variance) is lost during the process, as I'm sure you can imagine. But\n # you have to drop the dimension down to two, otherwise you wouldn't be able\n # to visualize a 2D decision surface / boundary. In the wild, you'd probably\n # leave in a lot more dimensions, which is better for higher accuracy, but\n # worse for visualizing the decision boundary;\n #\n # Your model should only be trained (fit) against the training data (data_train)\n # Once you've done this, you need use the model to transform both data_train\n # and data_test from their original high-D image feature space, down to 2D\n\n\n # TODO: Implement PCA here. ONLY train against your training data, but\n # transform both your training + test data, storing the results back into\n # data_train, and data_test.\n \n # .. your code here ..\n \n pca = PCA(n_components=2, svd_solver='full')\n pca.fit(x_train)\n\n data_train = pca.transform(x_train)\n data_test = pca.transform(x_test)\n\nelse:\n # INFO: Isomap is used *before* KNeighbors to simplify your high dimensionality\n # image samples down to just 2 components! A lot of information has been is\n # lost during the process, as I'm sure you can imagine. But if you have\n # non-linear data that can be represented on a 2D manifold, you probably will\n # be left with a far superior dataset to use for classification. Plus by\n # having the images in 2D space, you can plot them as well as visualize a 2D\n # decision surface / boundary. In the wild, you'd probably leave in a lot more\n # dimensions, which is better for higher accuracy, but worse for visualizing the\n # decision boundary;\n \n # Your model should only be trained (fit) against the training data (data_train)\n # Once you've done this, you need use the model to transform both data_train\n # and data_test from their original high-D image feature space, down to 2D\n\n \n # TODO: Implement Isomap here. ONLY train against your training data, but\n # transform both your training + test data, storing the results back into\n # data_train, and data_test.\n \n # .. your code here ..\n iso = manifold.Isomap(n_neighbors=5, n_components=2)\n iso.fit(x_train)\n data_train = iso.transform(x_train)\n data_test = iso.transform(x_test)", "_____no_output_____" ] ], [ [ "Implement `KNeighborsClassifier` here. You can use any K value from 1 through 20, so play around with it and attempt to get good accuracy. Fit the classifier against your training data and labels.", "_____no_output_____" ] ], [ [ "# .. your code here ..\nknn = KNeighborsClassifier(n_neighbors=3)\nknn = knn.fit(x_train, y_train)\nknn.score(x_test, y_test)", "_____no_output_____" ] ], [ [ "Calculate and display the accuracy of the testing set (data_test and label_test):", "_____no_output_____" ] ], [ [ "# .. your code here ..\nknn.score(x_test, y_test)\n\nscores = pd.DataFrame(columns=['n_neighbors', 'model_score'])\ntype(scores); scores.dtypes; scores.shape; scores.head(3)\nfor i in range(1, 21): # try K value from 1 through 20 in an attempt to find good accuracy\n score = KNeighborsClassifier(n_neighbors=i).fit(x_train, y_train).score(x_test, y_test)\n scores.loc[i-1] = [int(i), score] # or scores.loc[len(scores)] = [int(i), score]\n\nscores.model_score.unique() # | scores['model_score'] | scores.loc[:, 'model_score'] | scores.iloc[:, 1]\nscores.groupby('model_score').model_score.unique(); print(' ') # prints unique values with all decimal points\nscores.groupby('model_score').n_neighbors.unique(); print(' ')\nscores.groupby('model_score').n_neighbors.nunique()", " \n \n" ], [ "scores\n ", "_____no_output_____" ] ], [ [ "Let's chart the combined decision boundary, the training data as 2D plots, and the testing data as small images so we can visually validate performance:", "_____no_output_____" ] ], [ [ "Plot2DBoundary(x_train, x_train, y_test, y_test)", "_____no_output_____" ] ], [ [ "After submitting your answers, experiment with using using PCA instead of ISOMap. Are the results what you expected? Also try tinkering around with the test/train split percentage from 10-20%. Notice anything?", "_____no_output_____" ] ], [ [ "# .. your code changes above ..", "_____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", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "Unlicense" ]

[ "Unlicense" ]

[ "Unlicense" ]

[ [ [ "# PIC data", "_____no_output_____" ] ], [ [ "from astropy.constants import m_e, e, k_B\nk = k_B.value\nme = m_e.value\nq = e.value\nimport numpy as np\nimport matplotlib.pyplot as plt\n\nimport json\n%matplotlib notebook\n\nfrom scipy.interpolate import interp1d\nfrom math import ceil\nplt.style.use(\"presentation\")", "_____no_output_____" ], [ "with open(\"NewPic1D.dat\", \"r\") as f:\n dataPIC = json.load(f)\n# with open(\"PIC_data.dat\", \"r\") as f:\n# dataPIC = json.load(f)\n \nwith open(\"NewPIC_EVDFs.dat\", \"r\") as f:\n data = json.load(f)\n# with open(\"PIC_EVDFs.dat\", \"r\") as f:\n# data = json.load(f)\n \nprint(data.keys())\nprint(\"~~~~~~~~~~~~~~~ \\n\")\nprint(data[\"info\"])\nprint(\"~~~~~~~~~~~~~~~ \\n\")\nprint(\"Run disponibles\")\nfor k in [\"0\",\"1\",\"2\"]:\n run = data[k]\n print(k,\" p = \",run[\"p\"], \"mTorr\")\n \ndx = dataPIC[\"0\"][\"dx\"]", "dict_keys(['info', '0', '1', '2', '3', '4', '5'])\n~~~~~~~~~~~~~~~ \n\nInformations\np: pressur un mTorr\nprobnames: list of the names of the probes (also keys of the dict)\nprob_center: center of the prob\nin the prob data:\n position: the coords of the bin :ymin, ymax\n absciss: the velocities of the EVDFs (Vx, Vy, Vz)\n EVDF: the EVDFs (Vx, Vy, Vz)\n\n~~~~~~~~~~~~~~~ \n\nRun disponibles\n0 p = 0.1 mTorr\n1 p = 2.0 mTorr\n2 p = 10.0 mTorr\n" ], [ "k = '0'\nprobnames = np.array(data[k][\"probnames\"])\nprob_center = np.array(data[k][\"prob_center\"])\nprob_y0 = np.array(data[k][\"prob_y0\"])\nprob_y1 = np.array(data[k][\"prob_y1\"])\n\nprint(probnames)\nprint(prob_center)\ndx = data[k][\"dx\"]*1000\n", "['001' '002' '003' '004' '005' '006' '007' '008' '009' '010' '011' '012']\n[ 2.5 7.5 15. 25. 35. 50. 70. 90. 125. 175. 250. 450. ]\n" ], [ "def returnxy(pn, k=\"1\"):\n a = np.array(data[k][pn]['absciss'])\n V = np.array(data[k][pn]['EVDF'])\n idenx = 1\n x = a[:,idenx]\n x = x**2*np.sign(x)*me/q/2\n y = V[:,idenx]\n \n index = np.argwhere(pn == probnames)[0][0]\n xcenter = prob_center[index]\n x0 = int(prob_y0[index])\n x1 = int(prob_y1[index])\n \n phi = np.array(dataPIC[k][\"phi\"])\n pc = interp1d(np.arange(len(phi)),phi)(xcenter)\n p0 = phi[x0]\n p1 = phi[x1]\n \n # p = phi[int(xcenter)]\n return x, y, pc , p0, p1", "_____no_output_____" ], [ "# plot \nplt.figure(figsize=(4.5,4))\nplt.subplots_adjust(left=0.17, bottom=0.17, right=0.99, top=0.925, wspace=0.05, hspace=0.25)\n\nft = 14\ns = 2.5\n\nfor Nprob in range(len(probnames)):\n \n x, y, phic, phi0, phi1 = returnxy(probnames[Nprob])\n # x, y, phic = returnxy(probnames[Nprob])\n y0sum = (y).max()\n T = np.sum(np.abs(x) * y)/y.sum()*2\n plt.scatter(phic, T)\n \nphi = np.array(dataPIC[k][\"phi\"])\nTe = np.array(dataPIC[k][\"Te2\"])\nplt.plot(phi, Te,linewidth=s, alpha=0.7,ls=\"--\" )\n\n# plt.legend( fontsize=ft,loc=(1,0.1 ))\nplt.legend(loc = 'lower left', fontsize=11)\nplt.grid(alpha=0.5)\nplt.ylabel(\"Te\", fontsize=ft)\nplt.xlabel(\"$\\phi$ [V]\", fontsize=ft)\n\n", "/home/tavant/these/code/venv/stand/lib64/python3.7/site-packages/matplotlib/figure.py:2144: UserWarning: This figure was using constrained_layout==True, but that is incompatible with subplots_adjust and or tight_layout: setting constrained_layout==False. \n warnings.warn(\"This figure was using constrained_layout==True, \"\nNo handles with labels found to put in legend.\n" ] ], [ [ "# Heatflux from EVDF", "_____no_output_____" ] ], [ [ "k = \"0\"\nNprob = -1\nx, y, phic, phi0, phi1 = returnxy(probnames[Nprob], k=k)\nplt.figure(figsize=(4.5,4.5))\nplt.subplots_adjust(left=0.17, bottom=0.17, right=0.99, top=0.925, wspace=0.05, hspace=0.25)\n\nplt.plot(x,y)\n\nNprob = 1\nx, y, phic, phi0, phi1 = returnxy(probnames[Nprob], k=k)\nplt.plot(x,y)\n\nplt.yscale(\"log\")\nplt.vlines([phic,phic*1.3], 0.001,1e5)\nplt.ylim(bottom=10)", "_____no_output_____" ], [ "k = \"0\"\nNprob = 2\nx, y, phic, phi0, phi1 = returnxy(probnames[Nprob], k=k)\nplt.figure(figsize=(4.5,4.5))\nplt.subplots_adjust(left=0.17, bottom=0.17, right=0.99, top=0.925, wspace=0.05, hspace=0.25)\n\nplt.plot(x,y)\nplt.yscale(\"log\")\nplt.vlines([phic,phic*1.3], 0.001,1e5)\nplt.xlim(0, 20)\nplt.ylim(bottom=10)", "_____no_output_____" ], [ "from scipy.integrate import simps", "_____no_output_____" ], [ "def return_heat_flux(k=\"0\", Nprob=2, cut=True):\n \n x, y, phic, phi0, phi1 = returnxy(probnames[Nprob], k=k)\n y /= y.sum()\n if cut:\n mask = (x>phic) & (x<=1.1*phic)\n else:\n mask = (x>phic) \n \n \n heatflux = simps(0.5*x[mask]*y[mask], x[mask])\n \n flux = simps(y[mask], x[mask])\n \n x, y, phic, phi0, phi1 = returnxy(probnames[9], k=k)\n mask = (x>0) \n \n T = np.sum(np.abs(x[mask]) * y[mask])/y[mask].sum()*2 \n \n return heatflux/flux/T\n\nplt.figure()\n\nfor gamma, k in zip([1.6, 1.43, 1.41], [\"0\", \"1\", \"2\"]):\n \n plt.scatter(gamma, return_heat_flux(k, Nprob=3), c=\"k\", label=\"WITHOUT HIGH ENERGY TAIL\")\n plt.scatter(gamma, return_heat_flux(k, Nprob=3, cut=False), c=\"b\", label=\"WITH HIGH ENERGY TAIL\") \n \nplt.legend()", "_____no_output_____" ] ], [ [ "# conclusion\n\nWhen calculating the heat flux from the EVDF, we get that:\n$$\\frac{Q_e}{\\Gamma_e T_e} \\simeq 0.315$$\n\nWhich is very close from the theoritical value varing from 0.3 to 0.1.", "_____no_output_____" ] ] ]

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

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

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Capture Spectrum\n\nBelow we will generate plots for both the real and simulated capture spectra. If you are re-generating the plots, please be patient, as both load large amounts of data.\n\n## Measured Spectra\n\n*(If you are not interested in the code itself, you can collapse it by selecting the cell and then clicking on the bar to its left. You will still be able to run it and view its output.)*\n\nMeasured spectra and reconstructed PuBe rate. \nFirst, the reconstructed spetra were calculated by dividing the measured count rates by the livetimes and estimated effficiences after applying quality cuts. \nThen, the background rate was subtracted from the overall rate, leaving the events due to the PuBe source.\n\nA simulated measured spectrum was built using Geant4. \nWe then compared this to our actual data using MCMC sampling (`fitting` directory) as well as a simpler integral method (`Integral_Method.ipynb`).", "_____no_output_____" ] ], [ [ "#Import libraries and settings\nfrom IPython.core.display import display, HTML\ndisplay(HTML(\"<style>.container { width:100% !important; }</style>\"))\nexec(open(\"../python/nb_setup.py\").read())\nfrom constants import *\nimport R68_load as r68\nimport R68_efficiencies as eff\nmeas=r68.load_measured()\nimport R68_spec_tools as spec\nfrom matplotlib import *\nstyle.use('../mplstyles/stylelib/standard.mplstyle')\n\n#Turn off expected warnings\nlogging.getLogger('matplotlib.font_manager').disabled = True\nwarnings.filterwarnings(\"ignore\",category=RuntimeWarning)\n\nfig_w=7 #Used later for figure width\n\n# Binning setup\nEmax = 2000 #eVee\nEbins=np.linspace(0,Emax,201)\nEbins_ctr=(Ebins[:-1]+Ebins[1:])/2\nEfit_min=50#[eVee]\nEfit_max=1750#[eVee]\nspec_bounds=(np.digitize(Efit_min,Ebins)-1,np.digitize(Efit_max,Ebins)-1)\nEbins_ctr[slice(*spec_bounds)].shape\n\n#Measured\nN_meas_PuBe,_ = np.histogram(meas['PuBe']['E'],bins=Ebins)\nN_meas_Bkg,_ = np.histogram(meas['Bkg']['E'],bins=Ebins)\n#Count uncertainties are Poisson\ndN_meas_PuBe_Pois=np.sqrt(N_meas_PuBe)\ndN_meas_Bkg_Pois=np.sqrt(N_meas_Bkg)\n\n#Include uncertainty from efficiencies\ndN_meas_PuBe = N_meas_PuBe*np.sqrt( (dN_meas_PuBe_Pois/N_meas_PuBe)**2 + (eff.deff_write/eff.eff_write)**2 +\n (eff.dcutEffFit(Ebins_ctr)/eff.cutEffFit(Ebins_ctr))**2)\ndN_meas_Bkg = N_meas_Bkg*np.sqrt( (dN_meas_Bkg_Pois/N_meas_Bkg)**2 + (eff.deff_write_bkg/eff.eff_write_bkg)**2 +\n (eff.dcutEffFit_bkg(Ebins_ctr)/eff.cutEffFit_bkg(Ebins_ctr))**2)\n#Scaling factors\ntlive_ratio=meas['PuBe']['tlive']/meas['Bkg']['tlive']\nwriteEff_ratio=eff.eff_write/eff.eff_write_bkg\ndwriteEff_ratio=writeEff_ratio*np.sqrt( (eff.deff_write/eff.eff_write)**2 + (eff.deff_write_bkg/eff.eff_write_bkg)**2 )\n\ncutEff_ratio=eff.cutEffFit(Ebins_ctr)/eff.cutEffFit_bkg(Ebins_ctr)\ndcutEff_ratio = cutEff_ratio*np.sqrt( (eff.dcutEffFit(Ebins_ctr)/eff.cutEffFit(Ebins_ctr))**2 + \n (eff.dcutEffFit_bkg(Ebins_ctr)/eff.cutEffFit_bkg(Ebins_ctr))**2 )\n\nratio=tlive_ratio*writeEff_ratio*cutEff_ratio\ndratio=ratio*np.sqrt( (dwriteEff_ratio/writeEff_ratio)**2 +(dcutEff_ratio/cutEff_ratio)**2 )\n\n#Make sure any divide by 0s happened below threshold\nif (not np.all(np.isfinite(ratio[slice(*spec_bounds)]))) or (not np.all(np.isfinite(dratio[slice(*spec_bounds)]))):\n print('Error: Bad background scaling ratio in fit range.')\n\n#Bkg-subtracted measured PuBe signal\nN_bkg_scaled=N_meas_Bkg*ratio\ndN_bkg_scaled=N_bkg_scaled*np.sqrt( (dN_meas_Bkg/N_meas_Bkg)**2 + (dratio/ratio)**2 )\n\nN_meas = N_meas_PuBe - N_bkg_scaled\ndN_meas = np.sqrt( dN_meas_PuBe**2 + dN_bkg_scaled**2 ) #All errors are symmetric here when using the conservative cut eff fits\ndN_meas_stat = np.sqrt( dN_meas_PuBe_Pois**2 + (dN_meas_Bkg_Pois*ratio)**2 )\n\nDenom_PuBe = meas['PuBe']['tlive']*eff.eff_write*eff.cutEffFit(Ebins_ctr)*eff.trigEff(Ebins_ctr)\nR_meas_PuBe = N_meas_PuBe/Denom_PuBe\nDenom_Bkg = meas['Bkg']['tlive']*eff.eff_write_bkg*eff.cutEffFit_bkg(Ebins_ctr)*eff.trigEff(Ebins_ctr)\nR_meas_Bkg = N_meas_Bkg/Denom_Bkg\n\nR_meas = R_meas_PuBe - R_meas_Bkg\n\ndR_meas_stat_PuBe = R_meas_PuBe*dN_meas_PuBe_Pois/N_meas_PuBe\ndR_meas_PuBe = R_meas_PuBe*np.sqrt( (dN_meas_PuBe_Pois/N_meas_PuBe)**2 +\\\n (eff.deff_write/eff.eff_write)**2 +\\\n (eff.dcutEffFit(Ebins_ctr)/eff.cutEffFit(Ebins_ctr))**2 +\\\n (eff.dtrigEff(Ebins_ctr)/eff.trigEff(Ebins_ctr))**2 )\ndR_meas_stat_Bkg = R_meas_Bkg*dN_meas_Bkg_Pois/N_meas_Bkg\ndR_meas_Bkg = R_meas_Bkg*np.sqrt( (dN_meas_Bkg_Pois/N_meas_Bkg)**2 +\\\n (eff.deff_write_bkg/eff.eff_write_bkg)**2 +\\\n (eff.dcutEffFit_bkg(Ebins_ctr)/eff.cutEffFit_bkg(Ebins_ctr))**2 +\\\n (eff.dtrigEff(Ebins_ctr)/eff.trigEff(Ebins_ctr))**2 )\n\ndR_meas_stat = np.sqrt(dR_meas_stat_PuBe**2 + dR_meas_stat_Bkg**2)\ndR_meas = np.sqrt(dR_meas_PuBe**2 + dR_meas_Bkg**2)\n\nc_stat,dc_stat=spec.doBkgSub(r68.load_measured(verbose=False), Ebins, Efit_min, Efit_max, doEffsyst=False, doBurstLeaksyst=False, output='counts')\nc_syst,dc_syst=spec.doBkgSub(r68.load_measured(verbose=False), Ebins, Efit_min, Efit_max, doEffsyst=True, doBurstLeaksyst=False, output='counts')\nc_syst2,dc_syst2=spec.doBkgSub(r68.load_measured(verbose=False), Ebins, Efit_min, Efit_max, doEffsyst=True, doBurstLeaksyst=True, output='counts')\n\nr_stat,dr_stat=spec.doBkgSub(r68.load_measured(verbose=False), Ebins, Efit_min, Efit_max, doEffsyst=False, doBurstLeaksyst=False, output='reco-rate')\nr_syst,dr_syst=spec.doBkgSub(r68.load_measured(verbose=False), Ebins, Efit_min, Efit_max, doEffsyst=True, doBurstLeaksyst=False, output='reco-rate')\nr_syst2,dr_syst2=spec.doBkgSub(r68.load_measured(verbose=False), Ebins, Efit_min, Efit_max, doEffsyst=True, doBurstLeaksyst=True, output='reco-rate')\n\nfig,ax = plt.subplots(2,1,sharex=True,figsize=(16,16))\nfill_noise=ax[1].axvspan(0,50,color='r', alpha=0.25, label='No efficiency data',zorder=0)\n\ncthresh=Ebins_ctr>=50 #Only plot hists above threshold\n\n#Raw Count histograms\nax[0].set_title('Raw Counts, Statistical Uncertainties',size='40',pad='25')\n\nline_c_PuBe=ax[0].errorbar(Ebins_ctr,N_meas_PuBe,yerr=dN_meas_PuBe_Pois, drawstyle = 'steps-mid', linewidth=2)\nline_c_Bkg=ax[0].errorbar(Ebins_ctr,N_meas_Bkg,yerr=dN_meas_Bkg_Pois, drawstyle = 'steps-mid', linewidth=2)\nline_thresh=ax[0].axvline(50,linestyle='--',color='r', label='50 eV$_{ee}$ threshold',zorder=5)\n\nax[0].set_xlim(0,5e2)\n#ax[0].set_yscale('log')\nax[0].set_ylim(0,3e3)\nax[0].set_ylabel('Counts/bin')\n\n\nax[0].legend([line_c_PuBe, line_c_Bkg, line_thresh, fill_noise],\n [f\"PuBe, run time = {meas['PuBe']['tlive']/3600:.1f} hrs\",\n f\"Bkg, run time = {meas['Bkg']['tlive']/3600:.1f} hrs\",\n '50 eV$_{ee}$ threshold'],fontsize=30)\n \n#Reconstructed rate\n \n#Reverse errorbar order to be [lower,upper]\nax[1].set_title('Reconstructed Background-Subtracted Rate',size='40',pad='25')\nline_r_syst=ax[1].errorbar(Ebins_ctr[cthresh],60*r_syst2[cthresh],yerr=(60*dr_syst2[::-1])[:,cthresh],drawstyle = 'steps-mid', linewidth=2, label='Stat + Syst')\nline_r_stat=ax[1].errorbar(Ebins_ctr[cthresh],60*r_stat[cthresh],yerr=60*dr_stat[::-1][:,cthresh],drawstyle = 'steps-mid', linewidth=2, label='Stat')\n \nline_thresh=ax[1].axvline(50,linestyle='--',color='r', label='50 eV$_{ee}$ threshold',zorder=5)\nfill_noise=ax[1].axvspan(0,50,color='r', alpha=0.25, label='No efficiency data',zorder=0)\n \nax[1].legend([line_r_stat, line_r_syst, line_thresh, fill_noise],\n ['Stat', 'Stat + Syst', '50 eV$_{ee}$ threshold', 'No efficiency info'])\n \nax[1].set_ylim(0,3)\nax[1].set_xlim(0,500)\nax[1].set_ylabel('Counts/min/bin')\nax[1].set_xlabel('Energy [eV$_{ee}$]')\n\n\nplt.tight_layout()\nplt.savefig('../figures/meas_spec_reco_rate_pretty.pdf')\nplt.show()", "_____no_output_____" ] ], [ [ "Overlaid histograms comparingthe yielded energy PDFs for Sorensen and Lindhard models,\nincluding the resolution of the current detector (see `Calibration.ipynb`).\nThe histograms are comprised of approximately simulated cascades.\nThe orange (front) filled histogram represents the Lindhard model while the blue (back) filled histogram represents the Sorensen model.\nIn the Sorensen model, many points are pushed to zero due to the presenceof a cutoff energy,\nleading to a peak in the first bin that is not present in the Lindhard model.\nFor both models, we use k = 0.178, and for the Sorensen model, we use q= 0.00075.\nThe solid-line unfilled histogram represents the Lindhard model with one fifth of the detector’s resolution,\nand the dashed unfilled histogram represents the Sorensen model with one fifth of the detector’s resolution.\n\nThis is not the data generated by Geant4 mentioned above, \nand is intended only to give an idea of what simulated data looks like. \nThis data is instead generated using [a software package we developed for simulating nuclear recoils](https://github.com/villano-lab/nrCascadeSim).\nThis package does not assume any particular source and does not rely on the same assumptions as Geant4.", "_____no_output_____" ] ], [ [ "#Import Libraries\nimport uproot\nimport numpy as np\nimport matplotlib.pyplot as plt\nimport matplotlib.patches as mpatch\nplt.style.use('../mplstyles/stylelib/standard.mplstyle')\nfrom matplotlib.lines import Line2D\nfrom tabulate import tabulate\nimport sys\nsys.path.append('../python')\nimport nc_kinematics as nck\nimport lindhard as lin\nimport R68_yield as R68y\nfrom histogram_utils import histogramable as h\n\n#Select a file.\nfile = '../data/longsim.root'\n\n#It's running, there's just a lot of data. Please be patient!\n\nreal_Lind = np.ndarray.flatten(np.asarray(h(file)[0]))\nreal_Sor = np.ndarray.flatten(np.asarray(h(file,model='Sorenson')[0]))\nsmall_Lind = np.ndarray.flatten(np.asarray(h(file,scalefactor=0.2)[0]))\nsmall_Sor = np.ndarray.flatten(np.asarray(h(file,model='Sorenson',scalefactor=0.2)[0]))\n\nreal_Lind = real_Lind[real_Lind >= 0]\nreal_Sor = real_Sor[real_Sor >= 0]\nsmall_Lind = small_Lind[small_Lind >= 0]\nsmall_Sor = small_Sor[small_Sor >= 0]\n\ndef format_exponent(ax, axis='y'):\n\n # Change the ticklabel format to scientific format\n ax.ticklabel_format(axis=axis, style='sci', scilimits=(-2, 2))\n\n # Get the appropriate axis\n if axis == 'y':\n ax_axis = ax.yaxis\n x_pos = 0.0\n y_pos = 1.0\n horizontalalignment='left'\n verticalalignment='bottom'\n else:\n ax_axis = ax.xaxis\n x_pos = 1.0\n y_pos = -0.05\n horizontalalignment='right'\n verticalalignment='top'\n\n # Run plt.tight_layout() because otherwise the offset text doesn't update\n plt.tight_layout()\n\n # Get the offset value\n offset = ax_axis.get_offset_text().get_text()\n\n if len(offset) > 0:\n # Get that exponent value and change it into latex format\n minus_sign = u'\\u2212'\n expo = np.float(offset.replace(minus_sign, '-').split('e')[-1])\n offset_text = r'x$\\mathregular{10^{%d}}$' %expo\n\n # Turn off the offset text that's calculated automatically\n ax_axis.offsetText.set_visible(False)\n\n # Add in a text box at the top of the y axis\n ax.text(x_pos, y_pos, offset_text, transform=ax.transAxes,\n horizontalalignment=horizontalalignment,\n verticalalignment=verticalalignment,fontsize=30)\n return ax\n\nfig, ax = plt.subplots(figsize=(16,12))\n\nbinsize = 8 #bin width in eVee\nbins = np.arange(0,620,binsize)\n\nplt.hist(small_Lind,alpha=0.7,label='Small Res (1/5, Lindhard)',histtype='step',edgecolor='black',density='True',linewidth=2,bins=bins)\nplt.hist(small_Sor,alpha=0.7,label='Small Res (1/5, Sorenson)',histtype='step',edgecolor='black',linestyle='--',density='True',linewidth=2,bins=bins)\nplt.hist(real_Sor,alpha=0.6,label='Sorenson',histtype='step',fill=True,density='True',bins=bins,linewidth=3,edgecolor='navy',color='C0')\nplt.hist(real_Lind,alpha=0.6,label='Lindhard',histtype='step',fill=True,density='True',bins=bins,linewidth=3,edgecolor='#a30',color='C1')\n\nplt.xlabel(r\"Energy Yielded ($\\mathrm{eV}_{\\mathrm{ee}}$)\")\nplt.ylabel(\"Probability Distribution (eVee$^{-1}$)\")#Counts/(total counts * bin width)\")\n\nax = format_exponent(ax, axis='y')\nax.tick_params(axis='both',which='major')\n\nplt.xlim([0,None])\nplt.ylim([6e-13,6e-3]) #Make corner less awkward. Smallest starting value that will make the extra 0 go away\n\n#Legend\nLindPatch = mpatch.Patch(facecolor='C1',edgecolor='#a30',linewidth=3,label='Lindhard',alpha=0.6)\nSorPatch = mpatch.Patch(facecolor='C0',edgecolor='navy',linewidth=3,label='Sorenson',alpha=0.6)\nLindLine = Line2D([0],[0],alpha=0.7,color='black',label='Small Res (1/5, Lindhard)')\nSorLine = Line2D([0],[0],linestyle='--',alpha=0.7,color='black',label='Small Res (1/5, Sorenson)')\n\nplt.legend(handles=[LindPatch,SorPatch,LindLine,SorLine])\nplt.savefig('../figures/SorVsLin.pdf')\nplt.show()", "dict_keys(['xx', 'yy', 'ex', 'ey'])\n" ] ] ]

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

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

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "# Statistical exploration for Bayesian analysis of PhIP-seq", "_____no_output_____" ] ], [ [ "import pandas as pd\nimport numpy as np\nimport scipy as sp\nimport matplotlib.pyplot as plt\nimport seaborn as sns", "_____no_output_____" ], [ "%matplotlib inline", "_____no_output_____" ], [ "cpm = pd.read_csv('/Users/laserson/tmp/phip_analysis/phip-9/cpm.tsv', sep='\\t', header=0, index_col=0)", "_____no_output_____" ], [ "upper_bound = sp.stats.scoreatpercentile(cpm.values.ravel(), 99.9)\nupper_bound", "_____no_output_____" ], [ "fig, ax = plt.subplots()\n_ = ax.hist(cpm.values.ravel(), bins=100, log=True)\n_ = ax.set(title='cpm')", "_____no_output_____" ], [ "fig, ax = plt.subplots()\n_ = ax.hist(np.log10(cpm.values.ravel() + 0.5), bins=100, log=False)\n_ = ax.set(title='log10(cpm + 0.5)')", "_____no_output_____" ], [ "fig, ax = plt.subplots()\n_ = ax.hist(np.log10(cpm.values.ravel() + 0.5), bins=100, log=True)\n_ = ax.set(title='log10(cpm + 0.5)')", "_____no_output_____" ] ], [ [ "Plot only the lowest 99.9% of the data", "_____no_output_____" ] ], [ [ "fig, ax = plt.subplots()\n_ = ax.hist(cpm.values.ravel()[cpm.values.ravel() <= upper_bound], bins=range(100), log=False)\n_ = ax.set(xlim=(0, 60))\n_ = ax.set(title='trimmed cpm')", "_____no_output_____" ], [ "trimmed_cpm = cpm.values.ravel()[cpm.values.ravel() <= upper_bound]\ntrimmed_cpm.mean(), trimmed_cpm.std()", "_____no_output_____" ], [ "means = cpm.apply(lambda x: x[x <= upper_bound].mean(), axis=1, raw=True)\n_, edges = np.histogram(means, bins=[sp.stats.scoreatpercentile(means, p) for p in np.linspace(0, 100, 10)])", "_____no_output_____" ], [ "def plot_hist(ax, a):\n h, e = np.histogram(a, bins=100, range=(0, upper_bound), density=True)\n ax.hlines(h, e[:-1], e[1:])", "_____no_output_____" ], [ "for i in range(len(edges[:-1])):\n left = edges[i]\n right = edges[i + 1]\n rows = (means >= left) & (means <= right)\n values = cpm[rows].values.ravel()\n fig, ax = plt.subplots()\n plot_hist(ax, values)\n ax.set(xlim=(0, 50), title='mean in ({}, {})'.format(left, right))", "_____no_output_____" ] ], [ [ "Do the slices look Poisson?", "_____no_output_____" ] ], [ [ "a = np.random.poisson(8, 10000)\nfig, ax = plt.subplots()\nplot_hist(ax, a)\nax.set(xlim=(0, 50))", "_____no_output_____" ] ], [ [ "For the most part. Maybe try NegBin just in case", "_____no_output_____" ], [ "What does the distribution of the trimmed means look like?", "_____no_output_____" ] ], [ [ "fig, ax = plt.subplots()\nplot_hist(ax, means)\nax.set(xlim=(0, 50))", "_____no_output_____" ], [ "a = np.random.gamma(1, 10, 10000)\nfig, ax = plt.subplots()\nplot_hist(ax, a)\nax.set(xlim=(0, 50))", "_____no_output_____" ], [ "means.mean()", "_____no_output_____" ] ], [ [ "Following Anders and Huber, _Genome Biology_ 2010, compute some of their stats", "_____no_output_____" ], [ "Compute size factors", "_____no_output_____" ] ], [ [ "s = np.exp(np.median(np.log(cpm.values + 0.5) - np.log(cpm.values + 0.5).mean(axis=1).reshape((cpm.shape[0], 1)), axis=0))", "_____no_output_____" ], [ "_ = sns.distplot(s)", "_____no_output_____" ], [ "q = (cpm.values / s).mean(axis=1)", "_____no_output_____" ], [ "fig, ax = plt.subplots()\n_ = ax.hist(q, bins=100, log=False)", "_____no_output_____" ], [ "fig, ax = plt.subplots()\n_ = ax.hist(q, bins=100, log=True)", "_____no_output_____" ], [ "w = (cpm.values / s).std(axis=1, ddof=1)", "_____no_output_____" ], [ "fig, ax = plt.subplots()\n_ = ax.hist(w, bins=100, log=True)", "_____no_output_____" ], [ "fig, ax = plt.subplots()\n_ = ax.scatter(q, w)", "_____no_output_____" ], [ "_ = sns.lmplot('q', 'w', pd.DataFrame({'q': q, 'w': w}))", "_____no_output_____" ], [ "list(zip(cpm.values.sum(axis=0), s))", "_____no_output_____" ], [ "s", "_____no_output_____" ], [ "a = np.random.gamma(30, 1/30, 1000)", "_____no_output_____" ], [ "sns.distplot(a)", "_____no_output_____" ] ], [ [ "Proceeding with the following strategy/model\n\nTrim data to remove top 0.1% of count values. Compute mean of each row and use the means to fit a gamma distribution. Using these values, define a posterior on a rate for each clone, assuming Poisson stats for each cell. This means the posterior is also gamma distributed. Then compute the probability of seeing a more extreme value, weighted with the posterior on r_i.", "_____no_output_____" ] ], [ [ "import pystan", "_____no_output_____" ], [ "cpm = pd.read_csv('/Users/laserson/tmp/phip_analysis/phip-9/cpm.tsv', sep='\\t', header=0, index_col=0)", "_____no_output_____" ], [ "upper_bound = sp.stats.scoreatpercentile(cpm.values, 99.9)", "_____no_output_____" ], [ "trimmed_means = cpm.apply(lambda x: x[x <= upper_bound].mean(), axis=1, raw=True).values", "_____no_output_____" ], [ "brm = pystan.StanModel(model_name='background_rates', file='/Users/laserson/repos/bamophip/background_rates.stan')", "INFO:pystan:COMPILING THE C++ CODE FOR MODEL background_rates_0a7b2d07f26077b0d278330f6aedefc9 NOW.\n" ], [ "data = {\n 'num_clones': trimmed_means.shape[0],\n 'trimmed_means': trimmed_means\n}\nbr_fit = brm.sampling(data=data, iter=2000, chains=4)", "_____no_output_____" ], [ "br_fit", "_____no_output_____" ], [ "br_fit.plot()", "_____no_output_____" ], [ "alpha, beta, _ = br_fit.get_posterior_mean().mean(axis=1)", "_____no_output_____" ], [ "alpha, beta", "_____no_output_____" ], [ "h, e = np.histogram(np.random.gamma(alpha, 1 / beta, 50000), bins='auto', density=True)\nfig, ax = plt.subplots()\n_ = ax.hist(trimmed_means, bins=100, normed=True)\n_ = ax.hlines(h, e[:-1], e[1:])\n_ = ax.set(xlim=(0, 50))", "_____no_output_____" ], [ "# assumes the counts for each clone are Poisson distributed with the learned Gamma prior\n# Therefore, the posterior is Gamma distributed, and we use the expression for its expected value\ntrimmed_sums = cpm.apply(lambda x: x[x <= upper_bound].sum(), axis=1, raw=True).values\ntrimmed_sizes = cpm.apply(lambda x: (x <= upper_bound).sum(), axis=1, raw=True).values\nbackground_rates = (alpha + trimmed_sums) / (beta + trimmed_sizes)", "_____no_output_____" ], [ "# mlxp is \"minus log 10 pval\"\nmlxp = []\nfor i in range(cpm.shape[0]):\n mlxp.append(-sp.stats.poisson.logsf(cpm.values[i], background_rates[i]) / np.log(10))\nmlxp = np.asarray(mlxp)", "/Users/laserson/miniconda3/lib/python3.6/site-packages/scipy/stats/_distn_infrastructure.py:899: RuntimeWarning: divide by zero encountered in log\n return log(self._sf(x, *args))\n" ], [ "fig, ax = plt.subplots()\nh, e = np.histogram(10**(-mlxp.ravel()), bins='auto')\nax.hlines(h, e[:-1], e[1:])\nax.set(xlim=(0, 1))", "_____no_output_____" ], [ "fig, ax = plt.subplots()\nfinite = np.isfinite(mlxp.ravel())\n_ = ax.hist(mlxp.ravel()[finite], bins=100, log=True)", "_____no_output_____" ], [ "fig, ax = plt.subplots()\nfinite = np.isfinite(mlxp.ravel())\n_ = ax.hist(np.log10(mlxp.ravel()[finite] + 0.5), bins=100, log=True)", "_____no_output_____" ], [ "old_pvals = pd.read_csv('/Users/laserson/tmp/phip_analysis/phip-9/pvals.tsv', sep='\\t', header=0, index_col=0)", "_____no_output_____" ], [ "fig, ax = plt.subplots()\nh, e = np.histogram(10**(-old_pvals.values.ravel()), bins='auto')\nax.hlines(h, e[:-1], e[1:])\nax.set(xlim=(0, 1))", "_____no_output_____" ], [ "(old_pvals.values.ravel() > 10).sum()", "_____no_output_____" ], [ "(mlxp > 10).sum()", "_____no_output_____" ], [ "len(mlxp.ravel())", "_____no_output_____" ] ], [ [ "Can we use scipy's MLE for the gamma parameters instead?", "_____no_output_____" ] ], [ [ "sp.stats.gamma.fit(trimmed_means)", "_____no_output_____" ], [ "fig, ax = plt.subplots()\n_ = ax.hist(sp.stats.gamma.rvs(a=0.3387, loc=0, scale=3.102, size=10000), bins=100)\n_ = ax.set(xlim=(0, 50))", "_____no_output_____" ] ], [ [ "Hmmm...doesn't appear to get the correct solution.", "_____no_output_____" ], [ "Alternatively, let's try optimizing the log likelihood ourselves", "_____no_output_____" ] ], [ [ "pos = trimmed_means > 0\nn = len(trimmed_means)\ns = trimmed_means[pos].sum()\nsl = np.log(trimmed_means[pos]).sum()\ndef ll(x):\n return -1 * (n * x[0] * np.log(x[1]) - n * sp.special.gammaln(x[0]) + (x[0] - 1) * sl - x[1] * s)", "_____no_output_____" ], [ "param = sp.optimize.minimize(ll, np.asarray([2, 1]), bounds=[(np.nextafter(0, 1), None), (np.nextafter(0, 1), None)])\nparam", "_____no_output_____" ], [ "param.x", "_____no_output_____" ] ], [ [ "SUCCESS!", "_____no_output_____" ], [ "Do the p-values have a correlation with the peptide abundance?", "_____no_output_____" ] ], [ [ "mlxp = pd.read_csv('/Users/laserson/tmp/phip_analysis/sjogrens/mlxp.tsv', sep='\\t', index_col=0, header=0)\ninputs = pd.read_csv('/Users/laserson/repos/phage_libraries_private/human90/inputs/human90-larman1-input.tsv', sep='\\t', index_col=0, header=0)", "_____no_output_____" ], [ "m = pd.merge(mlxp, inputs, left_index=True, right_index=True)", "_____no_output_____" ], [ "sample = 'Sjogrens.serum.Sjogrens.FS12-03967.20A20G.1'", "_____no_output_____" ], [ "sp.stats.pearsonr(10**(-m[sample]), m['input'])", "_____no_output_____" ], [ "sp.stats.spearmanr(10**(-m[sample]), m['input'])", "_____no_output_____" ], [ "fig, ax = plt.subplots()\n_ = ax.scatter(10**(-m[sample]), m['input'])", "_____no_output_____" ], [ "fig, ax = plt.subplots()\n_ = ax.scatter(m[sample], m['input'])", "_____no_output_____" ], [ "h, xe, ye = np.histogram2d(m[sample], m['input'], bins=100)\nfig, ax = plt.subplots()\n_ = ax.imshow(h)", "_____no_output_____" ], [ "np.histogram2d", "_____no_output_____" ] ] ]

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

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "<a href=\"https://colab.research.google.com/github/evergreenllc2020/Deep-Learning-For-Hackers/blob/master/19_intent_classification.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>", "_____no_output_____" ], [ "# Intent Recognition with BERT using Keras and TensorFlow 2", "_____no_output_____" ] ], [ [ "!nvidia-smi", "Sun Mar 1 05:57:39 2020 \n+-----------------------------------------------------------------------------+\n| NVIDIA-SMI 440.48.02 Driver Version: 418.67 CUDA Version: 10.1 |\n|-------------------------------+----------------------+----------------------+\n| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |\n| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |\n|===============================+======================+======================|\n| 0 Tesla T4 Off | 00000000:00:04.0 Off | 0 |\n| N/A 71C P0 31W / 70W | 0MiB / 15079MiB | 0% Default |\n+-------------------------------+----------------------+----------------------+\n \n+-----------------------------------------------------------------------------+\n| Processes: GPU Memory |\n| GPU PID Type Process name Usage |\n|=============================================================================|\n| No running processes found |\n+-----------------------------------------------------------------------------+\n" ], [ "!pip install tensorflow-gpu >> /dev/null", "\u001b[31mERROR: tensorflow 1.15.0 has requirement tensorboard<1.16.0,>=1.15.0, but you'll have tensorboard 2.1.0 which is incompatible.\u001b[0m\n\u001b[31mERROR: tensorflow 1.15.0 has requirement tensorflow-estimator==1.15.1, but you'll have tensorflow-estimator 2.1.0 which is incompatible.\u001b[0m\n" ], [ "!pip install --upgrade grpcio >> /dev/null", "\u001b[31mERROR: tensorflow 1.15.0 has requirement tensorboard<1.16.0,>=1.15.0, but you'll have tensorboard 2.1.0 which is incompatible.\u001b[0m\n\u001b[31mERROR: tensorflow 1.15.0 has requirement tensorflow-estimator==1.15.1, but you'll have tensorflow-estimator 2.1.0 which is incompatible.\u001b[0m\n" ], [ "!pip install tqdm >> /dev/null", "_____no_output_____" ], [ "!pip install bert-for-tf2 >> /dev/null", "_____no_output_____" ], [ "!pip install sentencepiece >> /dev/null", "_____no_output_____" ], [ "import os\nimport math\nimport datetime\n\nfrom tqdm import tqdm\n\nimport pandas as pd\nimport numpy as np\n\nimport tensorflow as tf\nfrom tensorflow import keras\n\nimport bert\nfrom bert import BertModelLayer\nfrom bert.loader import StockBertConfig, map_stock_config_to_params, load_stock_weights\nfrom bert.tokenization.bert_tokenization import FullTokenizer\n\nimport seaborn as sns\nfrom pylab import rcParams\nimport matplotlib.pyplot as plt\nfrom matplotlib.ticker import MaxNLocator\nfrom matplotlib import rc\n\nfrom sklearn.metrics import confusion_matrix, classification_report\n\n%matplotlib inline\n%config InlineBackend.figure_format='retina'\n\nsns.set(style='whitegrid', palette='muted', font_scale=1.2)\n\nHAPPY_COLORS_PALETTE = [\"#01BEFE\", \"#FFDD00\", \"#FF7D00\", \"#FF006D\", \"#ADFF02\", \"#8F00FF\"]\n\nsns.set_palette(sns.color_palette(HAPPY_COLORS_PALETTE))\n\nrcParams['figure.figsize'] = 12, 8\n\nRANDOM_SEED = 42\n\nnp.random.seed(RANDOM_SEED)\ntf.random.set_seed(RANDOM_SEED)", "_____no_output_____" ] ], [ [ "# Data\n\nThe data contains various user queries categorized into seven intents. It is hosted on [GitHub](https://github.com/snipsco/nlu-benchmark/tree/master/2017-06-custom-intent-engines) and is first presented in [this paper](https://arxiv.org/abs/1805.10190).", "_____no_output_____" ] ], [ [ "!gdown --id 1OlcvGWReJMuyYQuOZm149vHWwPtlboR6 --output train.csv\n!gdown --id 1Oi5cRlTybuIF2Fl5Bfsr-KkqrXrdt77w --output valid.csv\n!gdown --id 1ep9H6-HvhB4utJRLVcLzieWNUSG3P_uF --output test.csv", "Downloading...\nFrom: https://drive.google.com/uc?id=1OlcvGWReJMuyYQuOZm149vHWwPtlboR6\nTo: /content/train.csv\n100% 799k/799k [00:00<00:00, 115MB/s]\nDownloading...\nFrom: https://drive.google.com/uc?id=1Oi5cRlTybuIF2Fl5Bfsr-KkqrXrdt77w\nTo: /content/valid.csv\n100% 43.3k/43.3k [00:00<00:00, 59.0MB/s]\nDownloading...\nFrom: https://drive.google.com/uc?id=1ep9H6-HvhB4utJRLVcLzieWNUSG3P_uF\nTo: /content/test.csv\n100% 43.1k/43.1k [00:00<00:00, 62.5MB/s]\n" ], [ "train = pd.read_csv(\"train.csv\")\nvalid = pd.read_csv(\"valid.csv\")\ntest = pd.read_csv(\"test.csv\")", "_____no_output_____" ], [ "train = train.append(valid).reset_index(drop=True)", "_____no_output_____" ], [ "train.shape", "_____no_output_____" ], [ "train.head()", "_____no_output_____" ], [ "chart = sns.countplot(train.intent, palette=HAPPY_COLORS_PALETTE)\nplt.title(\"Number of texts per intent\")\nchart.set_xticklabels(chart.get_xticklabels(), rotation=30, horizontalalignment='right');", "_____no_output_____" ] ], [ [ "# Intent Recognition with BERT", "_____no_output_____" ] ], [ [ "!wget https://storage.googleapis.com/bert_models/2018_10_18/uncased_L-12_H-768_A-12.zip", "--2020-03-01 05:59:09-- https://storage.googleapis.com/bert_models/2018_10_18/uncased_L-12_H-768_A-12.zip\nResolving storage.googleapis.com (storage.googleapis.com)... 172.217.214.128, 2607:f8b0:4001:c05::80\nConnecting to storage.googleapis.com (storage.googleapis.com)|172.217.214.128|:443... connected.\nHTTP request sent, awaiting response... 200 OK\nLength: 407727028 (389M) [application/zip]\nSaving to: ‘uncased_L-12_H-768_A-12.zip’\n\nuncased_L-12_H-768_ 100%[===================>] 388.84M 229MB/s in 1.7s \n\n2020-03-01 05:59:11 (229 MB/s) - ‘uncased_L-12_H-768_A-12.zip’ saved [407727028/407727028]\n\n" ], [ "!unzip uncased_L-12_H-768_A-12.zip", "Archive: uncased_L-12_H-768_A-12.zip\n creating: uncased_L-12_H-768_A-12/\n inflating: uncased_L-12_H-768_A-12/bert_model.ckpt.meta \n inflating: uncased_L-12_H-768_A-12/bert_model.ckpt.data-00000-of-00001 \n inflating: uncased_L-12_H-768_A-12/vocab.txt \n inflating: uncased_L-12_H-768_A-12/bert_model.ckpt.index \n inflating: uncased_L-12_H-768_A-12/bert_config.json \n" ], [ "os.makedirs(\"model\", exist_ok=True)", "_____no_output_____" ], [ "!mv uncased_L-12_H-768_A-12/ model", "_____no_output_____" ], [ "bert_model_name=\"uncased_L-12_H-768_A-12\"\n\nbert_ckpt_dir = os.path.join(\"model/\", bert_model_name)\nbert_ckpt_file = os.path.join(bert_ckpt_dir, \"bert_model.ckpt\")\nbert_config_file = os.path.join(bert_ckpt_dir, \"bert_config.json\")", "_____no_output_____" ] ], [ [ "## Preprocessing", "_____no_output_____" ] ], [ [ "class IntentDetectionData:\n DATA_COLUMN = \"text\"\n LABEL_COLUMN = \"intent\"\n\n def __init__(self, train, test, tokenizer: FullTokenizer, classes, max_seq_len=192):\n self.tokenizer = tokenizer\n self.max_seq_len = 0\n self.classes = classes\n \n ((self.train_x, self.train_y), (self.test_x, self.test_y)) = map(self._prepare, [train, test])\n\n print(\"max seq_len\", self.max_seq_len)\n self.max_seq_len = min(self.max_seq_len, max_seq_len)\n self.train_x, self.test_x = map(self._pad, [self.train_x, self.test_x])\n\n def _prepare(self, df):\n x, y = [], []\n \n for _, row in tqdm(df.iterrows()):\n text, label = row[IntentDetectionData.DATA_COLUMN], row[IntentDetectionData.LABEL_COLUMN]\n tokens = self.tokenizer.tokenize(text)\n tokens = [\"[CLS]\"] + tokens + [\"[SEP]\"]\n token_ids = self.tokenizer.convert_tokens_to_ids(tokens)\n self.max_seq_len = max(self.max_seq_len, len(token_ids))\n x.append(token_ids)\n y.append(self.classes.index(label))\n\n return np.array(x), np.array(y)\n\n def _pad(self, ids):\n x = []\n for input_ids in ids:\n input_ids = input_ids[:min(len(input_ids), self.max_seq_len - 2)]\n input_ids = input_ids + [0] * (self.max_seq_len - len(input_ids))\n x.append(np.array(input_ids))\n return np.array(x)", "_____no_output_____" ], [ "tokenizer = FullTokenizer(vocab_file=os.path.join(bert_ckpt_dir, \"vocab.txt\"))", "_____no_output_____" ], [ "tokenizer.tokenize(\"I can't wait to visit Bulgaria again!\")", "_____no_output_____" ], [ "tokens = tokenizer.tokenize(\"I can't wait to visit Bulgaria again!\")\ntokenizer.convert_tokens_to_ids(tokens)", "_____no_output_____" ], [ "def create_model(max_seq_len, bert_ckpt_file):\n\n with tf.io.gfile.GFile(bert_config_file, \"r\") as reader:\n bc = StockBertConfig.from_json_string(reader.read())\n bert_params = map_stock_config_to_params(bc)\n bert_params.adapter_size = None\n bert = BertModelLayer.from_params(bert_params, name=\"bert\")\n \n input_ids = keras.layers.Input(shape=(max_seq_len, ), dtype='int32', name=\"input_ids\")\n bert_output = bert(input_ids)\n\n print(\"bert shape\", bert_output.shape)\n\n cls_out = keras.layers.Lambda(lambda seq: seq[:, 0, :])(bert_output)\n cls_out = keras.layers.Dropout(0.5)(cls_out)\n logits = keras.layers.Dense(units=768, activation=\"tanh\")(cls_out)\n logits = keras.layers.Dropout(0.5)(logits)\n logits = keras.layers.Dense(units=len(classes), activation=\"softmax\")(logits)\n\n model = keras.Model(inputs=input_ids, outputs=logits)\n model.build(input_shape=(None, max_seq_len))\n\n load_stock_weights(bert, bert_ckpt_file)\n \n return model", "_____no_output_____" ] ], [ [ "## Training", "_____no_output_____" ] ], [ [ "classes = train.intent.unique().tolist()\n\ndata = IntentDetectionData(train, test, tokenizer, classes, max_seq_len=128)", "13784it [00:04, 3325.56it/s]\n700it [00:00, 3586.61it/s]\n" ], [ "data.train_x.shape", "_____no_output_____" ], [ "data.train_x[0]", "_____no_output_____" ], [ "data.train_y[0]", "_____no_output_____" ], [ "data.max_seq_len", "_____no_output_____" ], [ "model = create_model(data.max_seq_len, bert_ckpt_file)", "bert shape (None, 38, 768)\nDone loading 196 BERT weights from: model/uncased_L-12_H-768_A-12/bert_model.ckpt into <bert.model.BertModelLayer object at 0x7fa2d6676208> (prefix:bert). Count of weights not found in the checkpoint was: [0]. Count of weights with mismatched shape: [0]\nUnused weights from checkpoint: \n\tbert/embeddings/token_type_embeddings\n\tbert/pooler/dense/bias\n\tbert/pooler/dense/kernel\n\tcls/predictions/output_bias\n\tcls/predictions/transform/LayerNorm/beta\n\tcls/predictions/transform/LayerNorm/gamma\n\tcls/predictions/transform/dense/bias\n\tcls/predictions/transform/dense/kernel\n\tcls/seq_relationship/output_bias\n\tcls/seq_relationship/output_weights\n" ], [ "model.summary()", "Model: \"model\"\n_________________________________________________________________\nLayer (type) Output Shape Param # \n=================================================================\ninput_ids (InputLayer) [(None, 38)] 0 \n_________________________________________________________________\nbert (BertModelLayer) (None, 38, 768) 108890112 \n_________________________________________________________________\nlambda (Lambda) (None, 768) 0 \n_________________________________________________________________\ndropout (Dropout) (None, 768) 0 \n_________________________________________________________________\ndense (Dense) (None, 768) 590592 \n_________________________________________________________________\ndropout_1 (Dropout) (None, 768) 0 \n_________________________________________________________________\ndense_1 (Dense) (None, 7) 5383 \n=================================================================\nTotal params: 109,486,087\nTrainable params: 109,486,087\nNon-trainable params: 0\n_________________________________________________________________\n" ], [ "model.compile(\n optimizer=keras.optimizers.Adam(1e-5),\n loss=keras.losses.SparseCategoricalCrossentropy(from_logits=True),\n metrics=[keras.metrics.SparseCategoricalAccuracy(name=\"acc\")]\n)", "_____no_output_____" ], [ "log_dir = \"log/intent_detection/\" + datetime.datetime.now().strftime(\"%Y%m%d-%H%M%s\")\ntensorboard_callback = keras.callbacks.TensorBoard(log_dir=log_dir)\n\nhistory = model.fit(\n x=data.train_x, \n y=data.train_y,\n validation_split=0.1,\n batch_size=16,\n shuffle=True,\n epochs=5,\n callbacks=[tensorboard_callback]\n)", "Train on 12405 samples, validate on 1379 samples\nEpoch 1/5\n 5392/12405 [============>.................] - ETA: 2:51 - loss: 1.4535 - acc: 0.7361" ] ], [ [ "## Evaluation", "_____no_output_____" ] ], [ [ "%load_ext tensorboard", "_____no_output_____" ], [ "%tensorboard --logdir log", "_____no_output_____" ], [ "ax = plt.figure().gca()\nax.xaxis.set_major_locator(MaxNLocator(integer=True))\n\nax.plot(history.history['loss'])\nax.plot(history.history['val_loss'])\nplt.ylabel('Loss')\nplt.xlabel('Epoch')\nplt.legend(['train', 'test'])\nplt.title('Loss over training epochs')\nplt.show();", "_____no_output_____" ], [ "ax = plt.figure().gca()\nax.xaxis.set_major_locator(MaxNLocator(integer=True))\n\nax.plot(history.history['acc'])\nax.plot(history.history['val_acc'])\nplt.ylabel('Accuracy')\nplt.xlabel('Epoch')\nplt.legend(['train', 'test'])\nplt.title('Accuracy over training epochs')\nplt.show();", "_____no_output_____" ], [ "_, train_acc = model.evaluate(data.train_x, data.train_y)\n_, test_acc = model.evaluate(data.test_x, data.test_y)\n\nprint(\"train acc\", train_acc)\nprint(\"test acc\", test_acc)", "_____no_output_____" ], [ "y_pred = model.predict(data.test_x).argmax(axis=-1)", "_____no_output_____" ], [ "print(classification_report(data.test_y, y_pred, target_names=classes))", "_____no_output_____" ], [ "cm = confusion_matrix(data.test_y, y_pred)\ndf_cm = pd.DataFrame(cm, index=classes, columns=classes)", "_____no_output_____" ], [ "hmap = sns.heatmap(df_cm, annot=True, fmt=\"d\")\nhmap.yaxis.set_ticklabels(hmap.yaxis.get_ticklabels(), rotation=0, ha='right')\nhmap.xaxis.set_ticklabels(hmap.xaxis.get_ticklabels(), rotation=30, ha='right')\nplt.ylabel('True label')\nplt.xlabel('Predicted label');", "_____no_output_____" ], [ "sentences = [\n \"Play our song now\",\n \"Rate this book as awful\"\n]\n\npred_tokens = map(tokenizer.tokenize, sentences)\npred_tokens = map(lambda tok: [\"[CLS]\"] + tok + [\"[SEP]\"], pred_tokens)\npred_token_ids = list(map(tokenizer.convert_tokens_to_ids, pred_tokens))\n\npred_token_ids = map(lambda tids: tids +[0]*(data.max_seq_len-len(tids)),pred_token_ids)\npred_token_ids = np.array(list(pred_token_ids))\n\npredictions = model.predict(pred_token_ids).argmax(axis=-1)\n\nfor text, label in zip(sentences, predictions):\n print(\"text:\", text, \"\\nintent:\", classes[label])\n print()", "_____no_output_____" ] ], [ [ "# References\n\n- https://mccormickml.com/2019/07/22/BERT-fine-tuning/\n- https://github.com/snipsco/nlu-benchmark/tree/master/2017-06-custom-intent-engines\n- https://jalammar.github.io/illustrated-bert/\n- https://towardsdatascience.com/bert-for-dummies-step-by-step-tutorial-fb90890ffe03\n- https://www.reddit.com/r/MachineLearning/comments/ao23cp/p_how_to_use_bert_in_kaggle_competitions_a/", "_____no_output_____" ] ] ]

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

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

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Packages\nlibrary(apeglm)", "_____no_output_____" ], [ "library(vsn)", "_____no_output_____" ], [ "library(tidyverse)", "_____no_output_____" ], [ "library(DESeq2)", "_____no_output_____" ], [ "library(ggplot2)", "_____no_output_____" ], [ "library(hexbin)", "_____no_output_____" ], [ "library(BiocParallel)", "_____no_output_____" ], [ "library(magick)\nstr(magick::magick_config())", "_____no_output_____" ] ] ]

[ "code" ]

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

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "empty" ] ] ]

[ "empty" ]

[ [ "empty" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "# Welche Dateitypen gibt es?", "_____no_output_____" ], [ "## 1. Verbindung zur Datenbank\nEs wird eine Verbindung zur Neo4j-Datenbank aufgebaut.", "_____no_output_____" ] ], [ [ "import py2neo\n\ngraph = py2neo.Graph(bolt=True, host='localhost', user='neo4j', password='neo4j')", "_____no_output_____" ] ], [ [ "## 2. Cypher-Abfrage\nEs wird eine Abfrage an die Datenbank gestellt. Das Ergebnis wird in einem Dataframe (pandas) gespeichert.", "_____no_output_____" ] ], [ [ "import pandas as pd\n\nquery =\"MATCH (f:Git:File) RETURN f.relativePath as relativePath\"\ndf = pd.DataFrame(graph.run(query).data())\n", "_____no_output_____" ] ], [ [ "## 3. Datenaufbereitung\nZur Kontrolle werden die ersten fünf Zeilen des Ergebnisses der Abfrage als Tabelle ausgegeben.", "_____no_output_____" ] ], [ [ "df.head()", "_____no_output_____" ] ], [ [ "Der folgende Codeabschnitt extrahiert die verschiedenen Dateitypen entsprechend der Dateiendung und zählt deren Häufigkeit. Die Dateitypen werden in der Variable <font face=\"Courier\">datatype</font> und deren Häufigkeit in der Variable <font face=\"Courier\">frequency</font> gespeichert.", "_____no_output_____" ] ], [ [ "# Extrahiere Dateitypen aus Spalte des Dataframes.\ndatatypes = df['relativePath'].str.rsplit('.', 1).str[1]\n\n# Zähle die Anzahl der Dateitypen und bilde diese in einem Series-Objekt ab.\nseries = datatypes.value_counts()\n\n# Erzeuge zwei Listen aus dem Series-Objekt.\ndatatype = list(series.index)\nfrequency = list(series)\n\n# Erzeuge die Kategorie \"andere\", in der alle Dateitypen gesammelt werden, die weniger oder genau 20 mal auftauchen.\nandere = 0\nfor wert in frequency[:]:\n index = frequency.index(wert)\n if wert <= 20:\n andere += wert\n datatype.remove(datatype[index])\n frequency.remove(wert)\nfrequency.append(andere)\ndatatype.append(\"andere\")\n\nprint(frequency)\nprint(datatype)\n", "[1383, 80, 41, 36, 21, 126]\n['java', 'html', 'class', 'gif', 'txt', 'andere']\n" ] ], [ [ "## 4. Visualisierung\nDie Daten werden mittels eines Pie Charts visualisiert.\n\n", "_____no_output_____" ] ], [ [ "from IPython.display import display, HTML\n\nbase_html = \"\"\"\n<!DOCTYPE html>\n<html>\n <head>\n <script type=\"text/javascript\" src=\"http://kozea.github.com/pygal.js/javascripts/svg.jquery.js\"></script>\n <script type=\"text/javascript\" src=\"https://kozea.github.io/pygal.js/2.0.x/pygal-tooltips.min.js\"\"></script>\n </head>\n <body>\n <figure>\n {rendered_chart}\n </figure>\n </body>\n</html>\n\"\"\"", "_____no_output_____" ], [ "# Erstelle Pie Chart.\nimport pygal \npie_chart = pygal.Pie()\npie_chart.title = 'Dateitypen'\nfor einzelneDateitypen in datatype:\n index= datatype.index(einzelneDateitypen)\n anzahl=frequency[index]\n pie_chart.add(einzelneDateitypen, anzahl)\ndisplay(HTML(base_html.format(rendered_chart=pie_chart.render(is_unicode=True))))", "_____no_output_____" ] ] ]

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

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

[ "Unlicense" ]

[ "Unlicense" ]

[ "Unlicense" ]

[ [ [ "import pandas as pd \n\ndata = pd.read_csv(\"thanksgiving.csv\", encoding=\"Latin-1\")\n\n# print(data.head())\n\n# print(data.columns)\n\nPositive_survey_counts = data['Do you celebrate Thanksgiving?'].value_counts()\n\nPositive_surveys = data[data['Do you celebrate Thanksgiving?'] == 'Yes']\n\nMain_dish_counts = data['What is typically the main dish at your Thanksgiving dinner?'].value_counts()\n\nPositive_Tofurkey = data[data['What is typically the main dish at your Thanksgiving dinner?'] == 'Tofurkey']\n\nGravy_on_Tofurkey = Positive_Tofurkey['Do you typically have gravy?'].value_counts()\n\napple_isnull = data['Which type of pie is typically served at your Thanksgiving dinner? Please select all that apply. - Apple'].isnull()\npumpkin_isnull = data['Which type of pie is typically served at your Thanksgiving dinner? Please select all that apply. - Pumpkin'].isnull()\npecan_isnull = data['Which type of pie is typically served at your Thanksgiving dinner? Please select all that apply. - Pecan'].isnull()\n\nno_pies = apple_isnull & pumpkin_isnull & pecan_isnull\n\ndef string_to_income(string):\n if pd.isnull(string):\n return None\n string = string.split(' ')[0]\n if string == 'Prefer':\n return None\n else:\n string = string.split(' ')[0]\n string = string.replace('$','')\n string = string.replace(',','')\n return int(string)\n\ndata['int_income'] = data['How much total combined money did all members of your HOUSEHOLD earn last year?'].apply(string_to_income)\n \n# print(data['int_income'].describe())\n\ndef string_to_age(string):\n if pd.isnull(string):\n return None\n else: \n string = string.split(' ')[0]\n string = string.replace('+',\"\")\n return int(string)\n \ndata['int_age'] = data['Age'].apply(string_to_age)\n\n# print(data['int_age'].value_counts())\n# print(data['int_age'].describe())", "_____no_output_____" ], [ "low_income = data['int_income'] < 150000\nhigh_income = data['int_income'] > 150000\n\nlow_travel_results = data[low_income]['How far will you travel for Thanksgiving?'].value_counts(normalize = True)\nhigh_travel_results = data[high_income]['How far will you travel for Thanksgiving?'].value_counts(normalize = True) \n \nprint(low_travel_results)\nprint(high_travel_results)", "_____no_output_____" ], [ "So according to the results, there is a higher percentage of people who will \nhave Thanksgiving at home who make more that 150000.\n\nThere is a higher percentage of people who will travel for Thanksgiving in the \nless that 150000 bracket.", "_____no_output_____" ], [ "Friendsgiving1 = data.pivot_table(\n index = \"Have you ever tried to meet up with hometown friends on Thanksgiving night?\", \n columns = 'Have you ever attended a \"Friendsgiving?\"', \n values = \"int_age\")\n\nFriendsgiving2 = data.pivot_table(\n index = \"Have you ever tried to meet up with hometown friends on Thanksgiving night?\", \n columns = 'Have you ever attended a \"Friendsgiving?\"', \n values = \"int_income\")\nprint(Friendsgiving1)\nprint(Friendsgiving2)", "Have you ever attended a \"Friendsgiving?\" No Yes\nHave you ever tried to meet up with hometown fr... \nNo 42.283702 37.010526\nYes 41.475410 33.976744\nHave you ever attended a \"Friendsgiving?\" No Yes\nHave you ever tried to meet up with hometown fr... \nNo 78914.549654 72894.736842\nYes 78750.000000 66019.736842\n" ], [ "Looks like the younger and poorer you are, the more likely you are to attend a \nfriendsgiving. ", "_____no_output_____" ] ] ]

[ "code" ]

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

[ "CC0-1.0" ]

[ "CC0-1.0" ]

[ "CC0-1.0" ]

[ [ [ "## Introduction to the Interstellar Medium\n### Jonathan Williams", "_____no_output_____" ], [ "### Figure 6.3: portion of the Galactic plane in 21cm continuum showing bremsstrahlung and synchrotron sources", "_____no_output_____" ] ], [ [ "import numpy as np\nimport matplotlib.pyplot as plt\nimport matplotlib as mpl\nfrom astropy.io import fits\nfrom astropy.wcs import WCS\nfrom astropy.visualization import (ImageNormalize, SqrtStretch, LogStretch, AsinhStretch)\n%matplotlib inline", "_____no_output_____" ], [ "fig = plt.figure(figsize=(14,7.5))\n\nhdu = fits.open('g330to340.i.fits')\nwcs1 = WCS(hdu[0])\nax1 = fig.add_subplot(111, projection=wcs1)\nim1 = hdu[0].data\nhd1 = hdu[0].header\nhdu.close()\n#print(hd1)\n\nimin, imax = 380, 730\nimcrop = im1[:, imin:imax]\n#print(imcrop.min(),imcrop.max())\n\nnorm = ImageNormalize(imcrop, vmin=-0.15, vmax=2.0, stretch=AsinhStretch(a=0.1))\nax1.imshow(imcrop, cmap='gray', origin='lower', norm=norm) \n\nax1.set_xlim(0,350)\nax1.set_ylim(0,180)\nplt.plot([0,350], [90,90], ls='dashed', color='white', lw=2)\nax1.text(82, 45, 'HII', color='white', fontsize=18, fontweight='normal')\nax1.text(316, 97, 'SNR', color='white', fontsize=18, fontweight='normal')\n\n# scale bar\ndx = hd1['CDELT1']\n#print(dx)\n# 40'' per pixel, make bar 1 deg = 90 pix\nxbar = 90\nx0 = 250\nx1 = x0 + xbar\ny0 = 12\ndy = 2\nax1.plot([x0,x1],[y0,y0], 'w-', lw=2)\nax1.plot([x0,x0],[y0-dy,y0+dy], 'w-', lw=2)\nax1.plot([x1,x1],[y0-dy,y0+dy], 'w-', lw=2)\n\n# this crashes binder\n#mpl.rc('text', usetex=True)\n#mpl.rcParams['text.latex.preamble']=[r\"\\usepackage{amsmath}\"]\n#ax1.text(0.5*(x0+x1), y0+1.5*dy, r'$\\boldsymbol{1^\\circ}$', color='white', fontsize=24, fontweight='heavy', ha='center')\n\n# but this works ok\nax1.text(0.5*(x0+x1), y0+1.5*dy, r'$1^\\circ$', color='white', fontsize=24, fontweight='heavy', ha='center')\nax1.text(0.03,0.91,'21cm continuum', {'color': 'w', 'fontsize': 28}, transform=ax1.transAxes)\n\nfor i in (0,1):\n ax1.coords[i].set_ticks_visible(False)\n ax1.coords[i].set_ticklabel_visible(False)\n ax1.coords[i].set_ticks_visible(False)\n ax1.coords[i].set_ticklabel_visible(False)\n ax1.coords[i].set_axislabel('')\n ax1.coords[i].set_axislabel('')\n\nplt.tight_layout()\nplt.savefig('galactic_plane_continuum_21cm.pdf')", "-0.15332128 7.7325406\n" ] ] ]

[ "markdown", "code" ]

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

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# TV Script Generation\nIn this project, you'll generate your own [Simpsons](https://en.wikipedia.org/wiki/The_Simpsons) TV scripts using RNNs. You'll be using part of the [Simpsons dataset](https://www.kaggle.com/wcukierski/the-simpsons-by-the-data) of scripts from 27 seasons. The Neural Network you'll build will generate a new TV script for a scene at [Moe's Tavern](https://simpsonswiki.com/wiki/Moe's_Tavern).\n## Get the Data\nThe data is already provided for you. You'll be using a subset of the original dataset. It consists of only the scenes in Moe's Tavern. This doesn't include other versions of the tavern, like \"Moe's Cavern\", \"Flaming Moe's\", \"Uncle Moe's Family Feed-Bag\", etc..", "_____no_output_____" ] ], [ [ "\"\"\"\nDON'T MODIFY ANYTHING IN THIS CELL\n\"\"\"\nimport helper\n\ndata_dir = './data/simpsons/moes_tavern_lines.txt'\ntext = helper.load_data(data_dir)\n# Ignore notice, since we don't use it for analysing the data\ntext = text[81:]", "_____no_output_____" ] ], [ [ "## Explore the Data\nPlay around with `view_sentence_range` to view different parts of the data.", "_____no_output_____" ] ], [ [ "view_sentence_range = (0, 10)\n\n\"\"\"\nDON'T MODIFY ANYTHING IN THIS CELL\n\"\"\"\nimport numpy as np\n\nprint('Dataset Stats')\nprint('Roughly the number of unique words: {}'.format(len({word: None for word in text.split()})))\nscenes = text.split('\\n\\n')\nprint('Number of scenes: {}'.format(len(scenes)))\nsentence_count_scene = [scene.count('\\n') for scene in scenes]\nprint('Average number of sentences in each scene: {}'.format(np.average(sentence_count_scene)))\n\nsentences = [sentence for scene in scenes for sentence in scene.split('\\n')]\nprint('Number of lines: {}'.format(len(sentences)))\nword_count_sentence = [len(sentence.split()) for sentence in sentences]\nprint('Average number of words in each line: {}'.format(np.average(word_count_sentence)))\n\nprint()\nprint('The sentences {} to {}:'.format(*view_sentence_range))\nprint('\\n'.join(text.split('\\n')[view_sentence_range[0]:view_sentence_range[1]]))", "Dataset Stats\nRoughly the number of unique words: 11492\nNumber of scenes: 262\nAverage number of sentences in each scene: 15.251908396946565\nNumber of lines: 4258\nAverage number of words in each line: 11.50164396430249\n\nThe sentences 0 to 10:\n\nMoe_Szyslak: (INTO PHONE) Moe's Tavern. Where the elite meet to drink.\nBart_Simpson: Eh, yeah, hello, is Mike there? Last name, Rotch.\nMoe_Szyslak: (INTO PHONE) Hold on, I'll check. (TO BARFLIES) Mike Rotch. Mike Rotch. Hey, has anybody seen Mike Rotch, lately?\nMoe_Szyslak: (INTO PHONE) Listen you little puke. One of these days I'm gonna catch you, and I'm gonna carve my name on your back with an ice pick.\nMoe_Szyslak: What's the matter Homer? You're not your normal effervescent self.\nHomer_Simpson: I got my problems, Moe. Give me another one.\nMoe_Szyslak: Homer, hey, you should not drink to forget your problems.\nBarney_Gumble: Yeah, you should only drink to enhance your social skills.\n\n" ] ], [ [ "## Implement Preprocessing Functions\nThe first thing to do to any dataset is preprocessing. Implement the following preprocessing functions below:\n- Lookup Table\n- Tokenize Punctuation\n\n### Lookup Table\nTo create a word embedding, you first need to transform the words to ids. In this function, create two dictionaries:\n- Dictionary to go from the words to an id, we'll call `vocab_to_int`\n- Dictionary to go from the id to word, we'll call `int_to_vocab`\n\nReturn these dictionaries in the following tuple `(vocab_to_int, int_to_vocab)`", "_____no_output_____" ] ], [ [ "import numpy as np\nimport problem_unittests as tests\n\ndef create_lookup_tables(text):\n \"\"\"\n Create lookup tables for vocabulary\n :param text: The text of tv scripts split into words\n :return: A tuple of dicts (vocab_to_int, int_to_vocab)\n \"\"\"\n # TODO: Implement Function\n text = list(set(text))\n #text_id = range(len(text))\n #int_to_vocab = dict(zip(text_id, text)) \n #vocab_to_int = dict(zip(text, text_id))\n int_to_vocab = {word_i: word for word_i, word in enumerate(text)}\n vocab_to_int = {word: word_i for word_i, word in int_to_vocab.items()}\n return vocab_to_int, int_to_vocab\n\n\"\"\"\nDON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE\n\"\"\"\ntests.test_create_lookup_tables(create_lookup_tables)", "Tests Passed\n" ] ], [ [ "### Tokenize Punctuation\nWe'll be splitting the script into a word array using spaces as delimiters. However, punctuations like periods and exclamation marks make it hard for the neural network to distinguish between the word \"bye\" and \"bye!\".\n\nImplement the function `token_lookup` to return a dict that will be used to tokenize symbols like \"!\" into \"||Exclamation_Mark||\". Create a dictionary for the following symbols where the symbol is the key and value is the token:\n- Period ( . )\n- Comma ( , )\n- Quotation Mark ( \" )\n- Semicolon ( ; )\n- Exclamation mark ( ! )\n- Question mark ( ? )\n- Left Parentheses ( ( )\n- Right Parentheses ( ) )\n- Dash ( -- )\n- Return ( \\n )\n\nThis dictionary will be used to token the symbols and add the delimiter (space) around it. This separates the symbols as it's own word, making it easier for the neural network to predict on the next word. Make sure you don't use a token that could be confused as a word. Instead of using the token \"dash\", try using something like \"||dash||\".", "_____no_output_____" ] ], [ [ "def token_lookup():\n \"\"\"\n Generate a dict to turn punctuation into a token.\n :return: Tokenize dictionary where the key is the punctuation and the value is the token\n \"\"\"\n # TODO: Implement Function\n keys = ['.', ',', '\"', ';', '!', '?', '(', ')', '--','\\n']\n values = ['||Period||','||Comma||','||Quotation_Mark||','||Semicolon||','||Exclamation_mark||','||Question_mark||','||Left_Parentheses||','||Right_Parentheses||','||Dash||','||Return||']\n token_lookup = dict(zip(keys,values))\n return token_lookup\n\n\"\"\"\nDON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE\n\"\"\"\ntests.test_tokenize(token_lookup)", "Tests Passed\n" ] ], [ [ "## Preprocess all the data and save it\nRunning the code cell below will preprocess all the data and save it to file.", "_____no_output_____" ] ], [ [ "\"\"\"\nDON'T MODIFY ANYTHING IN THIS CELL\n\"\"\"\n# Preprocess Training, Validation, and Testing Data\nhelper.preprocess_and_save_data(data_dir, token_lookup, create_lookup_tables)", "_____no_output_____" ] ], [ [ "# Check Point\nThis is your first checkpoint. If you ever decide to come back to this notebook or have to restart the notebook, you can start from here. The preprocessed data has been saved to disk.", "_____no_output_____" ] ], [ [ "\"\"\"\nDON'T MODIFY ANYTHING IN THIS CELL\n\"\"\"\nimport helper\nimport numpy as np\nimport problem_unittests as tests\n\nint_text, vocab_to_int, int_to_vocab, token_dict = helper.load_preprocess()", "_____no_output_____" ] ], [ [ "## Build the Neural Network\nYou'll build the components necessary to build a RNN by implementing the following functions below:\n- get_inputs\n- get_init_cell\n- get_embed\n- build_rnn\n- build_nn\n- get_batches\n\n### Check the Version of TensorFlow and Access to GPU", "_____no_output_____" ] ], [ [ "\"\"\"\nDON'T MODIFY ANYTHING IN THIS CELL\n\"\"\"\nfrom distutils.version import LooseVersion\nimport warnings\nimport tensorflow as tf\n\n# Check TensorFlow Version\nassert LooseVersion(tf.__version__) >= LooseVersion('1.0'), 'Please use TensorFlow version 1.0 or newer'\nprint('TensorFlow Version: {}'.format(tf.__version__))\n\n# Check for a GPU\nif not tf.test.gpu_device_name():\n warnings.warn('No GPU found. Please use a GPU to train your neural network.')\nelse:\n print('Default GPU Device: {}'.format(tf.test.gpu_device_name()))", "TensorFlow Version: 1.0.0\n" ] ], [ [ "### Input\nImplement the `get_inputs()` function to create TF Placeholders for the Neural Network. It should create the following placeholders:\n- Input text placeholder named \"input\" using the [TF Placeholder](https://www.tensorflow.org/api_docs/python/tf/placeholder) `name` parameter.\n- Targets placeholder\n- Learning Rate placeholder\n\nReturn the placeholders in the following tuple `(Input, Targets, LearningRate)`", "_____no_output_____" ] ], [ [ "def get_inputs():\n \"\"\"\n Create TF Placeholders for input, targets, and learning rate.\n :return: Tuple (input, targets, learning rate)\n \"\"\"\n # TODO: Implement Function\n Input = tf.placeholder(dtype=tf.int32, shape=[None, None], name='input') \n Targets = tf.placeholder(dtype=tf.int32, shape=[None, None], name='targets') \n LearningRate = tf.placeholder(dtype=tf.float32, name='learning_rate') \n return Input, Targets, LearningRate\n\n\"\"\"\nDON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE\n\"\"\"\ntests.test_get_inputs(get_inputs)", "Tests Passed\n" ] ], [ [ "### Build RNN Cell and Initialize\nStack one or more [`BasicLSTMCells`](https://www.tensorflow.org/api_docs/python/tf/contrib/rnn/BasicLSTMCell) in a [`MultiRNNCell`](https://www.tensorflow.org/api_docs/python/tf/contrib/rnn/MultiRNNCell).\n- The Rnn size should be set using `rnn_size`\n- Initalize Cell State using the MultiRNNCell's [`zero_state()`](https://www.tensorflow.org/api_docs/python/tf/contrib/rnn/MultiRNNCell#zero_state) function\n - Apply the name \"initial_state\" to the initial state using [`tf.identity()`](https://www.tensorflow.org/api_docs/python/tf/identity)\n\nReturn the cell and initial state in the following tuple `(Cell, InitialState)`", "_____no_output_____" ] ], [ [ "def get_init_cell(batch_size, rnn_size):\n \"\"\"\n Create an RNN Cell and initialize it.\n :param batch_size: Size of batches\n :param rnn_size: Size of RNNs\n :return: Tuple (cell, initialize state)\n \"\"\"\n # TODO: Implement Function\n #rnn_layers = 2\n \n lstm = tf.contrib.rnn.BasicLSTMCell(rnn_size)\n Cell = tf.contrib.rnn.MultiRNNCell([lstm])\n #initial_state = Cell.zero_state(batch_size=tf.placeholder(dtype=tf.int32, shape=[]), dtype=tf.float32)\n InitialState = tf.identity(Cell.zero_state(batch_size, tf.float32), name = 'initial_state')\n #InitialState = tf.identity(initial_state, name='initial_state')\n return Cell, InitialState\n\n\"\"\"\nDON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE\n\"\"\"\ntests.test_get_init_cell(get_init_cell)", "Tests Passed\n" ] ], [ [ "### Word Embedding\nApply embedding to `input_data` using TensorFlow. Return the embedded sequence.", "_____no_output_____" ] ], [ [ "def get_embed(input_data, vocab_size, embed_dim):\n \"\"\"\n Create embedding for <input_data>.\n :param input_data: TF placeholder for text input.\n :param vocab_size: Number of words in vocabulary.\n :param embed_dim: Number of embedding dimensions\n :return: Embedded input.\n \"\"\"\n # TODO: Implement Function\n embedding = tf.Variable(tf.random_uniform((vocab_size, embed_dim), -1, 1))\n embed = tf.nn.embedding_lookup(embedding, input_data)\n return embed\n\n\n\"\"\"\nDON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE\n\"\"\"\ntests.test_get_embed(get_embed)", "Tests Passed\n" ] ], [ [ "### Build RNN\nYou created a RNN Cell in the `get_init_cell()` function. Time to use the cell to create a RNN.\n- Build the RNN using the [`tf.nn.dynamic_rnn()`](https://www.tensorflow.org/api_docs/python/tf/nn/dynamic_rnn)\n - Apply the name \"final_state\" to the final state using [`tf.identity()`](https://www.tensorflow.org/api_docs/python/tf/identity)\n\nReturn the outputs and final_state state in the following tuple `(Outputs, FinalState)` ", "_____no_output_____" ] ], [ [ "def build_rnn(cell, inputs):\n \"\"\"\n Create a RNN using a RNN Cell\n :param cell: RNN Cell\n :param inputs: Input text data\n :return: Tuple (Outputs, Final State)\n \"\"\"\n # TODO: Implement Function\n Outputs, Final_State = tf.nn.dynamic_rnn(cell, inputs, dtype=tf.float32)\n FinalState = tf.identity(Final_State, name='final_state')\n return Outputs, FinalState\n\n\"\"\"\nDON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE\n\"\"\"\ntests.test_build_rnn(build_rnn)", "Tests Passed\n" ] ], [ [ "### Build the Neural Network\nApply the functions you implemented above to:\n- Apply embedding to `input_data` using your `get_embed(input_data, vocab_size, embed_dim)` function.\n- Build RNN using `cell` and your `build_rnn(cell, inputs)` function.\n- Apply a fully connected layer with a linear activation and `vocab_size` as the number of outputs.\n\nReturn the logits and final state in the following tuple (Logits, FinalState) ", "_____no_output_____" ] ], [ [ "def build_nn(cell, rnn_size, input_data, vocab_size, embed_dim):\n \"\"\"\n Build part of the neural network\n :param cell: RNN cell\n :param rnn_size: Size of rnns\n :param input_data: Input data\n :param vocab_size: Vocabulary size\n :param embed_dim: Number of embedding dimensions\n :return: Tuple (Logits, FinalState)\n \"\"\"\n # TODO: Implement Function\n embedding = get_embed(input_data, vocab_size, embed_dim)\n Outputs, FinalState = build_rnn(cell, embedding)\n Logits = tf.contrib.layers.fully_connected(Outputs, vocab_size, activation_fn=None)\n return Logits, FinalState\n\n\"\"\"\nDON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE\n\"\"\"\ntests.test_build_nn(build_nn)", "Tests Passed\n" ] ], [ [ "### Batches\nImplement `get_batches` to create batches of input and targets using `int_text`. The batches should be a Numpy array with the shape `(number of batches, 2, batch size, sequence length)`. Each batch contains two elements:\n- The first element is a single batch of **input** with the shape `[batch size, sequence length]`\n- The second element is a single batch of **targets** with the shape `[batch size, sequence length]`\n\nIf you can't fill the last batch with enough data, drop the last batch.\n\nFor exmple, `get_batches([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20], 3, 2)` would return a Numpy array of the following:\n```\n[\n # First Batch\n [\n # Batch of Input\n [[ 1 2], [ 7 8], [13 14]]\n # Batch of targets\n [[ 2 3], [ 8 9], [14 15]]\n ]\n\n # Second Batch\n [\n # Batch of Input\n [[ 3 4], [ 9 10], [15 16]]\n # Batch of targets\n [[ 4 5], [10 11], [16 17]]\n ]\n\n # Third Batch\n [\n # Batch of Input\n [[ 5 6], [11 12], [17 18]]\n # Batch of targets\n [[ 6 7], [12 13], [18 1]]\n ]\n]\n```\n\nNotice that the last target value in the last batch is the first input value of the first batch. In this case, `1`. This is a common technique used when creating sequence batches, although it is rather unintuitive.", "_____no_output_____" ] ], [ [ "def get_batches(int_text, batch_size, seq_length):\n \"\"\"\n Return batches of input and target\n :param int_text: Text with the words replaced by their ids\n :param batch_size: The size of batch\n :param seq_length: The length of sequence\n :return: Batches as a Numpy array\n \"\"\"\n # TODO: Implement Function\n n_batches = len(int_text)//(batch_size*seq_length)\n input_batch = np.array(int_text[0: n_batches * batch_size * seq_length])\n target_batch = np.array(int_text[1: n_batches * batch_size * seq_length])\n target_batch = np.append(target_batch, int_text[0])\n \n input_batchs = np.split(input_batch.reshape(batch_size, -1), n_batches, 1)\n target_batchs = np.split(target_batch.reshape(batch_size, -1), n_batches, 1)\n \n get_batches = list(zip(input_batchs, target_batchs))\n return np.array(get_batches)\n\n\"\"\"\nDON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE\n\"\"\"\ntests.test_get_batches(get_batches)", "Tests Passed\n" ] ], [ [ "## Neural Network Training\n### Hyperparameters\nTune the following parameters:\n\n- Set `num_epochs` to the number of epochs.\n- Set `batch_size` to the batch size.\n- Set `rnn_size` to the size of the RNNs.\n- Set `embed_dim` to the size of the embedding.\n- Set `seq_length` to the length of sequence.\n- Set `learning_rate` to the learning rate.\n- Set `show_every_n_batches` to the number of batches the neural network should print progress.", "_____no_output_____" ] ], [ [ "# Number of Epochs\nnum_epochs = 100\n# Batch Size\nbatch_size = 156\n# RNN Size\nrnn_size = 600\n# Embedding Dimension Size\nembed_dim = 500\n# Sequence Length\nseq_length = 14\n# Learning Rate\nlearning_rate = 0.001\n# Show stats for every n number of batches\nshow_every_n_batches = 100\n\n\"\"\"\nDON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE\n\"\"\"\nsave_dir = './save'", "_____no_output_____" ] ], [ [ "### Build the Graph\nBuild the graph using the neural network you implemented.", "_____no_output_____" ] ], [ [ "\"\"\"\nDON'T MODIFY ANYTHING IN THIS CELL\n\"\"\"\nfrom tensorflow.contrib import seq2seq\n\ntrain_graph = tf.Graph()\nwith train_graph.as_default():\n vocab_size = len(int_to_vocab)\n input_text, targets, lr = get_inputs()\n input_data_shape = tf.shape(input_text)\n cell, initial_state = get_init_cell(input_data_shape[0], rnn_size)\n logits, final_state = build_nn(cell, rnn_size, input_text, vocab_size, embed_dim)\n\n # Probabilities for generating words\n probs = tf.nn.softmax(logits, name='probs')\n\n # Loss function\n cost = seq2seq.sequence_loss(\n logits,\n targets,\n tf.ones([input_data_shape[0], input_data_shape[1]]))\n\n # Optimizer\n optimizer = tf.train.AdamOptimizer(lr)\n\n # Gradient Clipping\n gradients = optimizer.compute_gradients(cost)\n capped_gradients = [(tf.clip_by_value(grad, -1., 1.), var) for grad, var in gradients if grad is not None]\n train_op = optimizer.apply_gradients(capped_gradients)", "_____no_output_____" ] ], [ [ "## Train\nTrain the neural network on the preprocessed data. If you have a hard time getting a good loss, check the [forms](https://discussions.udacity.com/) to see if anyone is having the same problem.", "_____no_output_____" ] ], [ [ "\"\"\"\nDON'T MODIFY ANYTHING IN THIS CELL\n\"\"\"\nbatches = get_batches(int_text, batch_size, seq_length)\n\nwith tf.Session(graph=train_graph) as sess:\n sess.run(tf.global_variables_initializer())\n\n for epoch_i in range(num_epochs):\n state = sess.run(initial_state, {input_text: batches[0][0]})\n\n for batch_i, (x, y) in enumerate(batches):\n feed = {\n input_text: x,\n targets: y,\n initial_state: state,\n lr: learning_rate}\n train_loss, state, _ = sess.run([cost, final_state, train_op], feed)\n\n # Show every <show_every_n_batches> batches\n if (epoch_i * len(batches) + batch_i) % show_every_n_batches == 0:\n print('Epoch {:>3} Batch {:>4}/{} train_loss = {:.3f}'.format(\n epoch_i,\n batch_i,\n len(batches),\n train_loss))\n\n # Save Model\n saver = tf.train.Saver()\n saver.save(sess, save_dir)\n print('Model Trained and Saved')", "Epoch 0 Batch 0/31 train_loss = 8.825\nEpoch 3 Batch 7/31 train_loss = 5.159\nEpoch 6 Batch 14/31 train_loss = 4.528\nEpoch 9 Batch 21/31 train_loss = 4.046\nEpoch 12 Batch 28/31 train_loss = 3.626\nEpoch 16 Batch 4/31 train_loss = 3.317\nEpoch 19 Batch 11/31 train_loss = 3.031\nEpoch 22 Batch 18/31 train_loss = 2.765\nEpoch 25 Batch 25/31 train_loss = 2.474\nEpoch 29 Batch 1/31 train_loss = 2.178\nEpoch 32 Batch 8/31 train_loss = 2.101\nEpoch 35 Batch 15/31 train_loss = 1.774\nEpoch 38 Batch 22/31 train_loss = 1.655\nEpoch 41 Batch 29/31 train_loss = 1.581\nEpoch 45 Batch 5/31 train_loss = 1.388\nEpoch 48 Batch 12/31 train_loss = 1.260\nEpoch 51 Batch 19/31 train_loss = 1.038\nEpoch 54 Batch 26/31 train_loss = 1.010\nEpoch 58 Batch 2/31 train_loss = 0.891\nEpoch 61 Batch 9/31 train_loss = 0.773\nEpoch 64 Batch 16/31 train_loss = 0.718\nEpoch 67 Batch 23/31 train_loss = 0.642\nEpoch 70 Batch 30/31 train_loss = 0.591\nEpoch 74 Batch 6/31 train_loss = 0.534\nEpoch 77 Batch 13/31 train_loss = 0.482\nEpoch 80 Batch 20/31 train_loss = 0.438\nEpoch 83 Batch 27/31 train_loss = 0.359\nEpoch 87 Batch 3/31 train_loss = 0.369\nEpoch 90 Batch 10/31 train_loss = 0.338\nEpoch 93 Batch 17/31 train_loss = 0.300\nEpoch 96 Batch 24/31 train_loss = 0.291\nModel Trained and Saved\n" ] ], [ [ "## Save Parameters\nSave `seq_length` and `save_dir` for generating a new TV script.", "_____no_output_____" ] ], [ [ "\"\"\"\nDON'T MODIFY ANYTHING IN THIS CELL\n\"\"\"\n# Save parameters for checkpoint\nhelper.save_params((seq_length, save_dir))", "_____no_output_____" ] ], [ [ "# Checkpoint", "_____no_output_____" ] ], [ [ "\"\"\"\nDON'T MODIFY ANYTHING IN THIS CELL\n\"\"\"\nimport tensorflow as tf\nimport numpy as np\nimport helper\nimport problem_unittests as tests\n\n_, vocab_to_int, int_to_vocab, token_dict = helper.load_preprocess()\nseq_length, load_dir = helper.load_params()", "_____no_output_____" ] ], [ [ "## Implement Generate Functions\n### Get Tensors\nGet tensors from `loaded_graph` using the function [`get_tensor_by_name()`](https://www.tensorflow.org/api_docs/python/tf/Graph#get_tensor_by_name). Get the tensors using the following names:\n- \"input:0\"\n- \"initial_state:0\"\n- \"final_state:0\"\n- \"probs:0\"\n\nReturn the tensors in the following tuple `(InputTensor, InitialStateTensor, FinalStateTensor, ProbsTensor)` ", "_____no_output_____" ] ], [ [ "def get_tensors(loaded_graph):\n \"\"\"\n Get input, initial state, final state, and probabilities tensor from <loaded_graph>\n :param loaded_graph: TensorFlow graph loaded from file\n :return: Tuple (InputTensor, InitialStateTensor, FinalStateTensor, ProbsTensor)\n \"\"\"\n # TODO: Implement Function\n with loaded_graph.as_default() as g:\n InputTensor = loaded_graph.get_tensor_by_name(\"input:0\")\n InitialStateTensor = loaded_graph.get_tensor_by_name(\"initial_state:0\")\n FinalStateTensor = loaded_graph.get_tensor_by_name(\"final_state:0\")\n ProbsTensor = loaded_graph.get_tensor_by_name(\"probs:0\")\n return InputTensor, InitialStateTensor, FinalStateTensor, ProbsTensor\n\n\n\"\"\"\nDON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE\n\"\"\"\ntests.test_get_tensors(get_tensors)", "Tests Passed\n" ] ], [ [ "### Choose Word\nImplement the `pick_word()` function to select the next word using `probabilities`.", "_____no_output_____" ] ], [ [ "def pick_word(probabilities, int_to_vocab):\n \"\"\"\n Pick the next word in the generated text\n :param probabilities: Probabilites of the next word\n :param int_to_vocab: Dictionary of word ids as the keys and words as the values\n :return: String of the predicted word\n \"\"\"\n # TODO: Implement Function\n pick_word = np.random.choice(len(int_to_vocab), 1, p=probabilities)[0]\n pick_word = int_to_vocab.get(pick_word)\n return pick_word\n\n\"\"\"\nDON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE\n\"\"\"\ntests.test_pick_word(pick_word)", "Tests Passed\n" ] ], [ [ "## Generate TV Script\nThis will generate the TV script for you. Set `gen_length` to the length of TV script you want to generate.", "_____no_output_____" ] ], [ [ "gen_length = 200\n# homer_simpson, moe_szyslak, or Barney_Gumble\nprime_word = 'moe_szyslak'\n\n\"\"\"\nDON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE\n\"\"\"\nloaded_graph = tf.Graph()\nwith tf.Session(graph=loaded_graph) as sess:\n # Load saved model\n loader = tf.train.import_meta_graph(load_dir + '.meta')\n loader.restore(sess, load_dir)\n\n # Get Tensors from loaded model\n input_text, initial_state, final_state, probs = get_tensors(loaded_graph)\n\n # Sentences generation setup\n gen_sentences = [prime_word + ':']\n prev_state = sess.run(initial_state, {input_text: np.array([[1]])})\n\n # Generate sentences\n for n in range(gen_length):\n # Dynamic Input\n dyn_input = [[vocab_to_int[word] for word in gen_sentences[-seq_length:]]]\n dyn_seq_length = len(dyn_input[0])\n\n # Get Prediction\n probabilities, prev_state = sess.run(\n [probs, final_state],\n {input_text: dyn_input, initial_state: prev_state})\n \n pred_word = pick_word(probabilities[dyn_seq_length-1], int_to_vocab)\n\n gen_sentences.append(pred_word)\n \n # Remove tokens\n tv_script = ' '.join(gen_sentences)\n for key, token in token_dict.items():\n ending = ' ' if key in ['\\n', '(', '\"'] else ''\n tv_script = tv_script.replace(' ' + token.lower(), key)\n tv_script = tv_script.replace('\\n ', '\\n')\n tv_script = tv_script.replace('( ', '(')\n \n print(tv_script)", "moe_szyslak: ah-ha, big mistake pal! hey moe, can you be the best book on you could never!\nhomer_simpson:(getting idea) but you're dea-d-d-dead.(three stooges scared sound)\ngrampa_simpson:(upbeat) i guess despite all sweet music, but then we pour it a beer at half something.\n\n\nlenny_leonard: hey, homer. r.\nhomer_simpson: moe, it's called!\nmoe_szyslak: guys, i'm gonna let him want to go to my dad\n\n\nhomer_simpson:(to moe) thirty cases of cough syrup. sign in the way.\nbarney_gumble: yeah, that's probably what i look at you, i'm too?\nmoe_szyslak: oh, here. the audience is still love over.\nmoe's_thoughts: this is kent brockman. and it begins,\" dear is to that!\nmoe_szyslak:(laughs) if you want to be back.\nvoice: excuse me, so you can either sit here in the back of my cruiser.\nhomer_simpson: well if i only got their secrets.\nlenny_leonard:(amiable) amanda\n" ] ], [ [ "# The TV Script is Nonsensical\nIt's ok if the TV script doesn't make any sense. We trained on less than a megabyte of text. In order to get good results, you'll have to use a smaller vocabulary or get more data. Luckly there's more data! As we mentioned in the begging of this project, this is a subset of [another dataset](https://www.kaggle.com/wcukierski/the-simpsons-by-the-data). We didn't have you train on all the data, because that would take too long. However, you are free to train your neural network on all the data. After you complete the project, of course.\n# Submitting This Project\nWhen submitting this project, make sure to run all the cells before saving the notebook. Save the notebook file as \"dlnd_tv_script_generation.ipynb\" and save it as a HTML file under \"File\" -> \"Download as\". Include the \"helper.py\" and \"problem_unittests.py\" files in your submission.", "_____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" ]

[ [ "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" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import numpy as np\nimport pandas as pd\nimport matplotlib.pyplot as plt\nimport matplotlib as mpl\nimport pandas_profiling as pp\nimport seaborn as sns\n\nfrom sklearn.model_selection import train_test_split\nfrom sklearn.preprocessing import OneHotEncoder\n\nfrom functions.preprocessing import Imputer, CategoricalEncoder, remove_outliers\n\nfrom lazypredict.Supervised import LazyRegressor\n\nplt.style.use('ggplot')", "_____no_output_____" ], [ "def remove_empty_features(data, threshold):\n \"\"\"...\"\"\"\n cols_to_drop = [column for column in data.columns \n if data[column].isna().mean() > threshold]\n data = data.drop(columns = cols_to_drop)\n return data, cols_to_drop\n\ndef mapping_from_list(order):\n return {label: idx for idx, label in enumerate(order)}\n\ndef ordinal_feature(data: pd.DataFrame, dictionary: dict):\n \"\"\" Transform ordinal features\n\n Args:\n data (dataframe)\n dictionary (dict)\n\n Returns:\n data (dataframe): encoded dataframe\n \"\"\"\n data_copy = data.copy()\n for key,value in dictionary.items():\n data_copy[key] = data_copy[key].map(mapping_from_list(value))\n\n return data_copy", "_____no_output_____" ], [ "# Road raw training data\nraw_data = pd.read_csv('train.csv', index_col=0)\nraw_data.head(5)", "_____no_output_____" ], [ "options = {\n \"MSSubClass\": {\"strategy\": \"most_frequent\"},\n \"MSZoning\": {\"strategy\": \"most_frequent\"},\n \"LotFrontage\": {\"strategy\": \"mean\"},\n \"LotArea\": {\"strategy\": \"mean\"},\n \"Street\": {\"strategy\": \"most_frequent\"},\n \"Alley\": {\"strategy\": \"constant\", \"fill_value\": \"NoAccess\"},\n \"LotShape\": {\"strategy\": \"most_frequent\"},\n \"LandContour\": {\"strategy\": \"most_frequent\"},\n \"Utilities\": {\"strategy\": \"most_frequent\"},\n \"LotConfig\": {\"strategy\": \"most_frequent\"},\n \"LandSlope\": {\"strategy\": \"most_frequent\"},\n \"Neighborhood\": {\"strategy\": \"most_frequent\"},\n \"Condition1\": {\"strategy\": \"most_frequent\"},\n \"Condition2\": {\"strategy\": \"most_frequent\"},\n \"Electrical\": {\"strategy\": \"most_frequent\"},\n \"1stFlrSF\": {\"strategy\": \"mean\"},\n \"2ndFlrSF\": {\"strategy\": \"mean\"},\n \"LowQualFinSF\": {\"strategy\": \"mean\"},\n \"GrLivArea\": {\"strategy\": \"mean\"},\n \"BsmtFullBath\": {\"strategy\": \"median\"},\n \"BsmtHalfBath\": {\"strategy\": \"median\"},\n \"FullBath\": {\"strategy\": \"median\"},\n \"HalfBath\": {\"strategy\": \"median\"},\n \"BedroomAbvGr\": {\"strategy\": \"median\"},\n \"KitchenAbvGr\": {\"strategy\": \"median\"},\n \"KitchenQual\": {\"strategy\": \"most_frequent\"},\n \"TotRmsAbvGrd\": {\"strategy\": \"median\"},\n \"BldgType\": {\"strategy\": \"most_frequent\"},\n \"HouseStyle\": {\"strategy\": \"most_frequent\"},\n \"OverallQual\": {\"strategy\": \"median\"},\n \"OverallCond\": {\"strategy\": \"median\"},\n \"YearBuilt\": {\"strategy\": \"median\"},\n \"YearRemodAdd\": {\"strategy\": \"median\"},\n \"RoofStyle\": {\"strategy\": \"most_frequent\"},\n \"RoofMatl\": {\"strategy\": \"most_frequent\"},\n \"Exterior1st\": {\"strategy\": \"most_frequent\"},\n \"Exterior2nd\": {\"strategy\": \"most_frequent\"},\n \"MasVnrType\": {\"strategy\": \"constant\", \"fill_value\": \"None\"},\n \"MasVnrArea\": {\"strategy\": \"mean\"},\n \"ExterQual\": {\"strategy\": \"most_frequent\"},\n \"ExterCond\": {\"strategy\": \"most_frequent\"},\n \"Foundation\": {\"strategy\": \"most_frequent\"},\n \"BsmtQual\": {\"strategy\": \"constant\", \"fill_value\": \"NoBasement\"},\n \"BsmtCond\": {\"strategy\": \"constant\", \"fill_value\": \"NoBasement\"},\n \"BsmtExposure\": {\"strategy\": \"constant\", \"fill_value\": \"NoBasement\"},\n \"BsmtFinType1\": {\"strategy\": \"constant\", \"fill_value\": \"NoBasement\"},\n \"BsmtFinSF1\": {\"strategy\": \"mean\"},\n \"BsmtFinType2\": {\"strategy\": \"constant\", \"fill_value\": \"NoBasement\"},\n \"BsmtFinSF2\": {\"strategy\": \"mean\"},\n \"BsmtUnfSF\": {\"strategy\": \"mean\"},\n \"TotalBsmtSF\": {\"strategy\": \"mean\"},\n \"Heating\": {\"strategy\": \"most_frequent\"},\n \"HeatingQC\": {\"strategy\": \"most_frequent\"},\n \"CentralAir\": {\"strategy\": \"most_frequent\"},\n \"ScreenPorch\": {\"strategy\": \"mean\"},\n \"PoolArea\": {\"strategy\": \"mean\"},\n \"PoolQC\": {\"strategy\": \"constant\", \"fill_value\": \"NoPool\"},\n \"Fence\": {\"strategy\": \"constant\", \"fill_value\": \"NoFence\"},\n \"MiscFeature\": {\"strategy\": \"constant\", \"fill_value\": \"None\"},\n \"MiscVal\": {\"strategy\": \"mean\"},\n \"MoSold\": {\"strategy\": \"median\"},\n \"YrSold\": {\"strategy\": \"median\"},\n \"SaleType\": {\"strategy\": \"most_frequent\"},\n \"SaleCondition\": {\"strategy\": \"most_frequent\"},\n \"Functional\": {\"strategy\": \"most_frequent\"},\n \"Fireplaces\": {\"strategy\": \"most_frequent\"},\n \"FireplaceQu\": {\"strategy\": \"constant\", \"fill_value\": \"NoAccess\"},\n \"GarageType\": {\"strategy\": \"constant\", \"fill_value\": \"NoAccess\"},\n \"GarageYrBlt\": {\"strategy\": \"most_frequent\"},\n \"GarageFinish\": {\"strategy\": \"constant\", \"fill_value\": \"NoAccess\"},\n \"GarageCars\": {\"strategy\": \"most_frequent\"},\n \"GarageArea\": {\"strategy\": \"median\"},\n \"GarageQual\": {\"strategy\": \"constant\", \"fill_value\": \"NoAccess\"},\n \"GarageCond\": {\"strategy\": \"constant\", \"fill_value\": \"NoAccess\"},\n \"PavedDrive\": {\"strategy\": \"most_frequent\"},\n \"WoodDeckSF\": {\"strategy\": \"most_frequent\"},\n \"OpenPorchSF\": {\"strategy\": \"most_frequent\"},\n \"EnclosedPorch\": {\"strategy\": \"mean\"},\n \"3SsnPorch\": {\"strategy\": \"most_frequent\"},\n}", "_____no_output_____" ], [ "params = {\n \"threshold_empty_features\": 0.3,\n}\n\ncols_to_drop = {\n \"remove_empty_features\": []\n}\n\ncategorical_colums = ['Exterior1st', 'Foundation', 'MasVnrType', 'Neighborhood', \n 'PavedDrive', 'Electrical', 'MSSubClass', 'SaleCondition',\n 'GarageType', 'Exterior2nd', 'MSZoning', 'CentralAir', \n 'Street','Alley','LandContour','Utilities','LotConfig', 'LandSlope', 'Condition1', 'Condition2', 'BldgType', 'HouseStyle', 'RoofStyle', 'RoofMatl', 'BsmtFinType2', 'Heating', 'Functional', 'GarageCond', 'Fence', 'MiscFeature', 'SaleType']\n\n# Ordinal features options\nordinal_columns = ['HeatingQC', 'GarageQual', 'BsmtFinType1', 'ExterQual', \n 'GarageFinish', 'BsmtExposure', 'LotShape', 'OverallQual',\n 'BsmtQual', 'KitchenQual']\n\nordinal_mapping = {\n 'BsmtExposure': ['None', 'No', 'Mn', 'Av', 'Gd'],\n 'BsmtFinType1': ['None', 'Unf', 'LwQ', 'Rec', 'BLQ', 'ALQ', 'GLQ'],\n 'GarageFinish': ['None', 'Unf', 'RFn', 'Fin'],\n 'LotShape': ['IR3', 'IR2', 'IR1', 'Reg']\n}\n\nordinal_common = ['ExterQual', 'ExterCond', 'BsmtQual', 'BsmtCond', 'HeatingQC',\n 'KitchenQual', 'FireplaceQu', 'GarageQual', 'PoolQC']\nfor column in ordinal_common:\n ordinal_mapping[column] = ['None', 'Po', 'Fa', 'TA', 'Gd']", "_____no_output_____" ], [ "# Removing features with a lot of missing values\ndata, cols_to_drop[\"remove_empty_features\"] = remove_empty_features(\n raw_data, \n params[\"threshold_empty_features\"]\n)\n\n# Impute missing values\nimp = Imputer(options=options)\ndata = imp.fit_transform(raw_data)\n\n# HOTFIX\nfor key in imp.options:\n if isinstance(imp.options[key]['_fill'], np.integer):\n imp.options[key]['_fill'] = int(imp.options[key]['_fill'])\nimp.save_options('imputer_options.json')\n\n# Encoding categorical features\nce = CategoricalEncoder(categorical_colums)\ndata = ce.fit_transform(data)\n\n# Encoding ordinal features\ndata = ordinal_feature(data, ordinal_mapping)\n\n# data\ndata", "_____no_output_____" ] ], [ [ "## Model metrics before removing outliers", "_____no_output_____" ] ], [ [ "reg = LazyRegressor()\nX = data.drop(columns = [\"SalePrice\"])\ny = data[\"SalePrice\"]\nX_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.3, random_state=42)\nmodels, _ = reg.fit(X_train, X_test, y_train, y_test)\nmodels", "100%|██████████| 43/43 [00:36<00:00, 1.19it/s]\n" ] ], [ [ "## Removing outliers", "_____no_output_____" ] ], [ [ "nan_columns = {column: data[column].isna().sum() for column in data.columns if data[column].isna().sum() > 0}\nnan_columns", "_____no_output_____" ], [ "data[\"PoolQC\"].sample(10)", "_____no_output_____" ], [ "ordinal_common = ['ExterQual', 'ExterCond', 'BsmtQual', 'BsmtCond', 'HeatingQC',\n 'KitchenQual', 'FireplaceQu', 'GarageQual', 'PoolQC']", "_____no_output_____" ], [ "outlier_removed_data = remove_outliers(data_no_empty_features, method=\"IsolationForest\", threshold=0.1, model_kwargs = {})", "Model to detect outliers is IsolationForest with parameters {}\n" ], [ "reg = LazyRegressor()\nX = outlier_removed_data.drop(columns = [\"SalePrice\"])\ny = outlier_removed_data[\"SalePrice\"]\nX_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=42)\nmodels, _ = reg.fit(X_train, X_test, y_train, y_test)\nmodels", "_____no_output_____" ] ], [ [ "## TODO:\n\n- Krzysiek:\n - funkcje zwracają indeksy i kolumny\n \n- kbdev\n - Encoding ordinal features as a class\n - fix np.int64 bug in json serialization\n - \n \n- miri\n - nie będzie jej (na 50%)\n \n- Patryk\n - zapis do pliku Encoder, konstruktor z pliku\n - PR \n \n```python\nour_encoder = OurOneHotEncoder(columns=...)\ndata = our_encoder.fit(data)\nour_encoder.save(file.json)\n \nour_encoder.from_file(file.json)\nour_encoder.transform(other_data)\n```\n", "_____no_output_____" ] ] ]

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

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

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "<a href=\"https://colab.research.google.com/github/Lore8614/Lore8614.github.io/blob/master/Big_Dreams.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", "_____no_output_____" ], [ "from google.colab import drive\ndrive.mount('/content/drive')", "Mounted at /content/drive\n" ], [ "%pwd\n%ls '/content/drive/My Drive/Machine Learning Final'", "ls: cannot access '/content/drive/My Drive/Machine Learning Final': No such file or directory\n" ], [ "pos_muts = pd.read_csv('/content/drive/My Drive/Machine Learning Final/H77_metadata.csv')\nfreqs = pd.read_csv('/content/drive/My Drive/Machine Learning Final/HCV1a_TsMutFreq_195.csv')\nmut_rate = pd.read_csv('/content/drive/My Drive/Machine Learning Final/Geller.mutation.rates_update.csv')\nfreqs.head()\n", "_____no_output_____" ], [ "mut_rate.head()", "_____no_output_____" ], [ "pos_muts.head()", "_____no_output_____" ], [ "# Start Calculating costs", "_____no_output_____" ] ] ]

[ "markdown", "code" ]

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

[ "CC-BY-4.0" ]

[ "CC-BY-4.0" ]

[ "CC-BY-4.0" ]

[ [ [ "# Nested cross-validation\n\nIn this notebook, we show a pattern called **nested cross-validation** which\nshould be used when you want to both evaluate a model and tune the\nmodel's hyperparameters.\n\nCross-validation is a powerful tool to evaluate the statistical performance\nof a model. It is also used to select the best model from a pool of models.\nThis pool of models can be the same family of predictor but with different\nparameters. In this case, we call this procedure **hyperparameter tuning**.\n\nWe could also imagine that we would like to choose among heterogeneous models\nthat will similarly use the cross-validation.\n\nBefore we go into details regarding the nested cross-validation, we will\nfirst recall the pattern used to fine tune a model's hyperparameters.\n\nLet's load the breast cancer dataset.", "_____no_output_____" ] ], [ [ "from sklearn.datasets import load_breast_cancer\n\ndata, target = load_breast_cancer(return_X_y=True)", "_____no_output_____" ] ], [ [ "Now, we'll make a minimal example using the utility `GridSearchCV` to find\nthe best parameters via cross-validation.", "_____no_output_____" ] ], [ [ "from sklearn.model_selection import GridSearchCV\nfrom sklearn.svm import SVC\n\nparam_grid = {\"C\": [0.1, 1, 10], \"gamma\": [.01, .1]}\nmodel_to_tune = SVC()\n\nsearch = GridSearchCV(estimator=model_to_tune, param_grid=param_grid,\n n_jobs=2)\nsearch.fit(data, target)", "_____no_output_____" ] ], [ [ "We recall that `GridSearchCV` will train a model with some specific parameter\non a training set and evaluate it on testing. However, this evaluation is\ndone via cross-validation using the `cv` parameter. This procedure is\nrepeated for all possible combinations of parameters given in `param_grid`.\n\nThe attribute `best_params_` will give us the best set of parameters that\nmaximize the mean score on the internal test sets.", "_____no_output_____" ] ], [ [ "print(f\"The best parameter found are: {search.best_params_}\")", "_____no_output_____" ] ], [ [ "We can now show the mean score obtained using the parameter `best_score_`.", "_____no_output_____" ] ], [ [ "print(f\"The mean score in CV is: {search.best_score_:.3f}\")", "_____no_output_____" ] ], [ [ "At this stage, one should be extremely careful using this score. The\nmisinterpretation would be the following: since the score was computed on a\ntest set, it could be considered our model's testing score.\n\nHowever, we should not forget that we used this score to pick-up the best\nmodel. It means that we used knowledge from the test set (i.e. test score) to\ndecide our model's training parameter.\n\nThus, this score is not a reasonable estimate of our testing error.\nIndeed, we can show that it will be too optimistic in practice. The good way\nis to use a \"nested\" cross-validation. We will use an inner cross-validation\ncorresponding to the previous procedure shown to optimize the\nhyperparameters. We will also include this procedure within an outer\ncross-validation, which will be used to estimate the testing error of\nour tuned model.\n\nIn this case, our inner cross-validation will always get the training set of\nthe outer cross-validation, making it possible to compute the testing\nscore on a completely independent set.\n\nWe will show below how we can create such nested cross-validation and obtain\nthe testing score.", "_____no_output_____" ] ], [ [ "from sklearn.model_selection import cross_val_score, KFold\n\n# Declare the inner and outer cross-validation\ninner_cv = KFold(n_splits=4, shuffle=True, random_state=0)\nouter_cv = KFold(n_splits=4, shuffle=True, random_state=0)\n\n# Inner cross-validation for parameter search\nmodel = GridSearchCV(\n estimator=model_to_tune, param_grid=param_grid, cv=inner_cv, n_jobs=2)\n\n# Outer cross-validation to compute the testing score\ntest_score = cross_val_score(model, data, target, cv=outer_cv, n_jobs=2)\nprint(f\"The mean score using nested cross-validation is: \"\n f\"{test_score.mean():.3f} +/- {test_score.std():.3f}\")", "_____no_output_____" ] ], [ [ "In the example above, the reported score is more trustful and should be close\nto production's expected statistical performance.\n\nWe will illustrate the difference between the nested and non-nested\ncross-validation scores to show that the latter one will be too optimistic in\npractice. In this regard, we will repeat several time the experiment and\nshuffle the data differently. Besides, we will store the score obtain with\nand without the nested cross-validation.", "_____no_output_____" ] ], [ [ "test_score_not_nested = []\ntest_score_nested = []\n\nN_TRIALS = 20\nfor i in range(N_TRIALS):\n inner_cv = KFold(n_splits=4, shuffle=True, random_state=i)\n outer_cv = KFold(n_splits=4, shuffle=True, random_state=i)\n\n # Non_nested parameter search and scoring\n model = GridSearchCV(estimator=model_to_tune, param_grid=param_grid,\n cv=inner_cv, n_jobs=2)\n model.fit(data, target)\n test_score_not_nested.append(model.best_score_)\n\n # Nested CV with parameter optimization\n test_score = cross_val_score(model, data, target, cv=outer_cv, n_jobs=2)\n test_score_nested.append(test_score.mean())", "_____no_output_____" ] ], [ [ "We can merge the data together and make a box plot of the two strategies.", "_____no_output_____" ] ], [ [ "import pandas as pd\n\nall_scores = {\n \"Not nested CV\": test_score_not_nested,\n \"Nested CV\": test_score_nested,\n}\nall_scores = pd.DataFrame(all_scores)", "_____no_output_____" ], [ "import matplotlib.pyplot as plt\n\ncolor = {\"whiskers\": \"black\", \"medians\": \"black\", \"caps\": \"black\"}\nall_scores.plot.box(color=color, vert=False)\nplt.xlabel(\"Accuracy\")\n_ = plt.title(\"Comparison of mean accuracy obtained on the test sets with\\n\"\n \"and without nested cross-validation\")", "_____no_output_____" ] ], [ [ "We observe that the model's statistical performance with the nested\ncross-validation is not as good as the non-nested cross-validation.\n\nAs a conclusion, when optimizing parts of the machine learning pipeline (e.g.\nhyperparameter, transform, etc.), one needs to use nested cross-validation to\nevaluate the statistical performance of the predictive model. Otherwise, the\nresults obtained without nested cross-validation are over-optimistic.", "_____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", "code" ], [ "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import matplotlib.pyplot as plt\nimport numpy as np\nimport emachine as EM\nimport itertools\nfrom joblib import Parallel, delayed\n#from sklearn.model_selection import train_test_split", "_____no_output_____" ], [ "np.random.seed(0)", "_____no_output_____" ], [ "n_var = 20 ; g = 0.5 ; n_seq = 5000", "_____no_output_____" ], [ "# Synthetic data are generated by using `generate_seq`.\nw_true,seqs = EM.generate_seq(n_var,n_seq,g=g)\nprint(seqs.shape)\n\nops = EM.operators(seqs)\nprint(ops.shape)", "(5000, 20)\n(5000, 210)\n" ], [ "# predict interactions w\neps_list = np.linspace(0.1,0.9,9)\nn_eps = len(eps_list)\nres = Parallel(n_jobs = n_eps)(delayed(EM.fit)(ops,eps=eps,max_iter=100) for eps in eps_list)\nw_eps = np.array([res[i][0] for i in range(len(res))])\nw_eps_iter = np.array([res[i][1] for i in range(len(res))])\n\n#e_eps = np.zeros(len(eps_list))\n#w_eps = np.zeros((len(eps_list),ops.shape[1]))\n#for i,eps in enumerate(eps_list):\n# w_eps[i,:],e_eps[i] = EM.fit(ops,w_true,eps=eps,max_iter=100)\n #print('eps and e_eps:',eps,e_eps[i])", "_____no_output_____" ], [ "w_eps_iter.shape", "_____no_output_____" ], [ "MSE = ((w_true[np.newaxis,np.newaxis,:] - w_eps_iter)**2).mean(axis=2)\nMSE.shape", "_____no_output_____" ], [ "# Entropy\n#w_iter_eps[n_eps,n_iter,n_ops]\n#ops[n_seq,n_ops] \nenergy_eps_iter = -np.sum((ops[:,np.newaxis,np.newaxis,:]*w_eps_iter[np.newaxis,:,:,:]),axis=3)\nprob_eps_iter = np.exp(energy_eps_iter) # [n_seq,n_eps,n_iter]\nprob_eps_iter /= prob_eps_iter.sum(axis=0)[np.newaxis,:,:] \nentropy_eps_iter = -(prob_eps_iter*np.log(prob_eps_iter)).sum(axis=0) #[n_eps,n_iter] ", "_____no_output_____" ], [ "entropy_eps_iter.shape", "_____no_output_____" ], [ "ieps_show = [2,4,8]\n\nnx,ny = 2,2\nfig, ax = plt.subplots(ny,nx,figsize=(nx*3.5,ny*3))\n\nfor i in ieps_show:\n ax[0,0].plot(MSE[i],label='eps=%1.1f'%eps_list[i])\n ax[1,0].plot(entropy_eps_iter[i,:],label='eps=%1.1f'%eps_list[i])\n\nax[0,1].plot(eps_list,MSE[:,-1],'ko-')\n\nax[1,1].plot(eps_list,entropy_eps_iter[:,-1],'ko-',label='final')\nax[1,1].plot(eps_list,entropy_eps_iter[:,:].max(axis=1),'r^--',label='max')\n\nax[0,0].legend()\nax[1,0].legend()\nax[1,1].legend()\n\nax[0,0].set_ylabel('MSE')\nax[0,1].set_ylabel('MSE')\nax[1,0].set_ylabel('Entropy')\nax[1,1].set_ylabel('Entropy')\n\nax[0,0].set_xlabel('Iterations')\nax[0,1].set_xlabel('epsilon')\nax[1,0].set_xlabel('Iterations')\nax[1,1].set_xlabel('epsilon')\n\nplt.tight_layout(h_pad=1, w_pad=1.5)\n#plt.savefig('fig.pdf', format='pdf', dpi=100)", "_____no_output_____" ] ] ]

[ "code" ]

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

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import warnings\nwarnings.filterwarnings('ignore')\n\nimport glob\nimport numpy as np\n\nfrom photutils import Background2D, SExtractorBackground\nfrom photutils import DAOStarFinder\nfrom photutils import CircularAperture,aperture_photometry\nfrom photutils.utils import calc_total_error\n\n\nimport astropy.wcs as wcs\nfrom astropy.io import fits\nfrom astropy.stats import sigma_clipped_stats, SigmaClip\nfrom astropy.nddata.utils import Cutout2D\nfrom astropy import units as u\n\nimport matplotlib.pyplot as plt\nfrom astropy.visualization import ZScaleInterval", "_____no_output_____" ], [ "mypath = '/Users/jielaizhang/Desktop/waissya/Testdata_Target_E/ASASJ030015-0459.7_20190913_B/'\noutmoviename='mymovie.gif'", "_____no_output_____" ], [ "# Load in all of the fits images in the directory and sort it\n\nimage_list = glob.glob(mypath+'*e91.fits.fz')\nimage_list.sort()", "_____no_output_____" ], [ "#Make some useful lists of values to track/record\n\nobstime = []\nBmag = []\nVmag = []\nBmag_e = []\nVmag_e = []\navg_offset = []", "_____no_output_____" ], [ "# Input the information for the calibration stars identified\n# in the previous notebook for batch processing of all of the images\n\nzpt_instrumental = 25.\n\ntar_ra = 45.064\ntar_dec = -4.995\ntar_color = 'yellow'\nref_ra = [44.93200, 45.00766, 45.11216, 45.12369]\nref_dec = [-5.03533, -4.79669, -4.91007, -4.93852]\nref_colors = ['red','cyan', 'green', 'blue']\nref_mag = [11.275, 12.093, 13.005, 14.65]", "_____no_output_____" ], [ "def do_phot_get_mag(data,hdr,err,ra,dec):\n positions = []\n zpt_instrumental = 25.\n w = wcs.WCS(hdr)\n xcoords, ycoords = w.all_world2pix(ra,dec,1)\n positions = np.transpose((xcoords, ycoords))\n apertures = CircularAperture(positions, r=24.)\n phot = aperture_photometry(data, apertures, error=err)\n\n mag = list(-2.5*np.log10(phot['aperture_sum']) + zpt_instrumental)\n dmag = list((2.5/np.log(10))*(phot['aperture_sum_err']/phot['aperture_sum']))\n \n return mag,dmag", "_____no_output_____" ], [ "def make_cutout(data,hdr,ra,dec):\n\n w = wcs.WCS(hdr)\n xcoord, ycoord = w.all_world2pix(ra,dec,1)\n position = np.transpose((xcoord, ycoord))\n size = u.Quantity([120, 120], u.pixel)\n cutout = Cutout2D(data, position, size, wcs=w, mode='strict')\n\n cutout_wcs = cutout.wcs\n header = cutout_wcs.to_header()\n hdu = fits.PrimaryHDU(data=cutout.data, header=header)\n\n return hdu", "_____no_output_____" ], [ "# Let's calculate the star's mag for *each* frame in the dataset\n\nfor frame in image_list:\n hdu = fits.open(frame)\n\n # Grab the actual science data based on above.\n sci_data = hdu[1]\n sci_hdr = sci_data.header\n time = sci_hdr['MJD-OBS']\n obstime.append(time)\n\n # Background estimation:\n sigma_clip = SigmaClip(sigma=3.) # Sigma clip bright obvious things to avoid biasing the background estimate\n bkg_estimator = SExtractorBackground() # Apply the SExtractor algorithm to our estimation\n bkg = Background2D(\n sci_data.data, (50, 50),\n filter_size=(3, 3),\n sigma_clip=sigma_clip,\n bkg_estimator=bkg_estimator)\n\n # Now let's subtract the background from the data\n sci_bkg = sci_data.data - bkg.background\n\n # Define an error image that will be used when calculating photometry\n effective_gain = 1.\n error = calc_total_error(sci_bkg, bkg.background_rms, effective_gain)\n\n # Calculate instrumental mags for each of the reference stars\n cal_mag,cal_dmag = do_phot_get_mag(sci_bkg,sci_hdr,error,ref_ra,ref_dec)\n\n # Calculate offsets and the standard deviation of the offset from each star.\n offsets = []\n for i in range(len(cal_mag)):\n offsets.append(ref_mag[i] - cal_mag[i])\n offset = np.mean(offsets)\n avg_offset.append(offset)\n doffset = np.std(offsets)\n \n # Do photometry on the variable target!!\n tar_mag,tar_dmag = do_phot_get_mag(sci_bkg,sci_hdr,error,tar_ra,tar_dec)\n \n cal_tar_mag = tar_mag[0]+offset\n cal_tar_dmag = np.sqrt(tar_dmag[0]**2.+doffset**2.)\n \n Bmag.append(cal_tar_mag)\n Bmag_e.append(cal_tar_dmag)\n \n # Make tiny cutouts of the variable star in each frame\n cutout_hdu = make_cutout(sci_bkg,sci_hdr,tar_ra,tar_dec)\n #cutout_hdu.writeto(frame+'_cutout.fits', overwrite=True)\n \n # Plot figures using these cutouts and output images\n interval = ZScaleInterval()\n vmin = interval.get_limits(cutout_hdu.data)[0]\n vmax = interval.get_limits(cutout_hdu.data)[1]\n\n plt.subplot(projection=wcs.WCS(cutout_hdu.header))\n plt.imshow(cutout_hdu.data, vmin=vmin, vmax=vmax, origin='lower')\n plt.xlabel('R.A.')\n plt.ylabel('Declination')\n \n pngname = str(time).replace('.','')\n plt.savefig(mypath+pngname+'.png', overwrite=True)\n \nBmag = np.array(Bmag)", "_____no_output_____" ], [ "# # Make a rudimentary lightcurve\n\nplt.figure(figsize=(10.5, 7))\nplt.errorbar(obstime,Bmag,xerr=None,yerr=Bmag_e, fmt='mo', capsize=9.0)\nplt.xlabel('MJD', fontsize=18)\nplt.ylabel('B Magnitude', fontsize=18)\nplt.show()", "_____no_output_____" ], [ "# Here we are going to use the cutouts we made above to make\n# an little movie of the variable star target changing brightness\n# over time and loop it!\n\nimport imageio\n\ncutout_list = glob.glob(mypath+'*.png')\ncutout_list.sort()\n\ncutout_frames = []\nfor file in cutout_list:\n cutout_frames.append(imageio.imread(file))\nimageio.mimsave(mypath+outmoviename, cutout_frames)", "_____no_output_____" ], [ "print(obstime)\nprint(list(Bmag))\nprint(Bmag_e)", "_____no_output_____" ] ] ]

[ "code" ]

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

[ "Apache-2.0", "CC-BY-4.0" ]

[ "Apache-2.0", "CC-BY-4.0" ]

[ "Apache-2.0", "CC-BY-4.0" ]

[ [ [ "%%html\n<link href=\"http://mathbook.pugetsound.edu/beta/mathbook-content.css\" rel=\"stylesheet\" type=\"text/css\" />\n<link href=\"https://aimath.org/mathbook/mathbook-add-on.css\" rel=\"stylesheet\" type=\"text/css\" />\n<style>.subtitle {font-size:medium; display:block}</style>\n<link href=\"https://fonts.googleapis.com/css?family=Open+Sans:400,400italic,600,600italic\" rel=\"stylesheet\" type=\"text/css\" />\n<link href=\"https://fonts.googleapis.com/css?family=Inconsolata:400,700&subset=latin,latin-ext\" rel=\"stylesheet\" type=\"text/css\" /><!-- Hide this cell. -->\n<script>\nvar cell = $(\".container .cell\").eq(0), ia = cell.find(\".input_area\")\nif (cell.find(\".toggle-button\").length == 0) {\nia.after(\n $('<button class=\"toggle-button\">Toggle hidden code</button>').click(\n function (){ ia.toggle() }\n )\n )\nia.hide()\n}\n</script>\n", "_____no_output_____" ] ], [ [ "**Important:** to view this notebook properly you will need to execute the cell above, which assumes you have an Internet connection. It should already be selected, or place your cursor anywhere above to select. Then press the \"Run\" button in the menu bar above (the right-pointing arrowhead), or press Shift-Enter on your keyboard.", "_____no_output_____" ], [ "$\\newcommand{\\identity}{\\mathrm{id}}\n\\newcommand{\\notdivide}{\\nmid}\n\\newcommand{\\notsubset}{\\not\\subset}\n\\newcommand{\\lcm}{\\operatorname{lcm}}\n\\newcommand{\\gf}{\\operatorname{GF}}\n\\newcommand{\\inn}{\\operatorname{Inn}}\n\\newcommand{\\aut}{\\operatorname{Aut}}\n\\newcommand{\\Hom}{\\operatorname{Hom}}\n\\newcommand{\\cis}{\\operatorname{cis}}\n\\newcommand{\\chr}{\\operatorname{char}}\n\\newcommand{\\Null}{\\operatorname{Null}}\n\\newcommand{\\lt}{<}\n\\newcommand{\\gt}{>}\n\\newcommand{\\amp}{&}\n$", "_____no_output_____" ], [ "<div class=\"mathbook-content\"><h2 class=\"heading hide-type\" alt=\"Section 8.2 Linear Codes\"><span class=\"type\">Section</span><span class=\"codenumber\">8.2</span><span class=\"title\">Linear Codes</span></h2><a href=\"section-linear-codes.ipynb\" class=\"permalink\">¶</a></div>", "_____no_output_____" ], [ "<div class=\"mathbook-content\"></div>", "_____no_output_____" ], [ "<div class=\"mathbook-content\"><p id=\"p-1232\">To gain more knowledge of a particular code and develop more efficient techniques of encoding, decoding, and error detection, we need to add additional structure to our codes. One way to accomplish this is to require that the code also be a group. A <dfn class=\"terminology\">group code</dfn> is a code that is also a subgroup of ${\\mathbb Z}_2^n\\text{.}$</p></div>", "_____no_output_____" ], [ "<div class=\"mathbook-content\"><p id=\"p-1233\">To check that a code is a group code, we need only verify one thing. If we add any two elements in the code, the result must be an $n$-tuple that is again in the code. It is not necessary to check that the inverse of the $n$-tuple is in the code, since every codeword is its own inverse, nor is it necessary to check that ${\\mathbf 0}$ is a codeword. For instance,</p><div class=\"displaymath\">\n\\begin{equation*}\n(11000101) + (11000101) = (00000000).\n\\end{equation*}\n</div></div>", "_____no_output_____" ], [ "<div class=\"mathbook-content\"><article class=\"example-like\" id=\"example-algcodes-weights\"><h6 class=\"heading\"><span class=\"type\">Example</span><span class=\"codenumber\">8.16</span></h6><p id=\"p-1234\">Suppose that we have a code that consists of the following 7-tuples:</p><div class=\"displaymath\">\n\\begin{align*}\n&(0000000) & & (0001111) & & (0010101) & & (0011010)\\\\\n&(0100110) & & (0101001) & & (0110011) & & (0111100)\\\\\n&(1000011) & & (1001100) & & (1010110) & & (1011001)\\\\\n&(1100101) & & (1101010) & & (1110000) & & (1111111).\n\\end{align*}\n</div><p>It is a straightforward though tedious task to verify that this code is also a subgroup of ${\\mathbb Z}_2^7$ and, therefore, a group code. This code is a single error-detecting and single error-correcting code, but it is a long and tedious process to compute all of the distances between pairs of codewords to determine that $d_{\\min} = 3\\text{.}$ It is much easier to see that the minimum weight of all the nonzero codewords is 3. As we will soon see, this is no coincidence. However, the relationship between weights and distances in a particular code is heavily dependent on the fact that the code is a group.</p></article></div>", "_____no_output_____" ], [ "<div class=\"mathbook-content\"><article class=\"theorem-like\" id=\"lemma-5\"><h6 class=\"heading\"><span class=\"type\">Lemma</span><span class=\"codenumber\">8.17</span></h6><p id=\"p-1235\">Let ${\\mathbf x}$ and ${\\mathbf y}$ be binary $n$-tuples. Then $w({\\mathbf x} + {\\mathbf y}) = d({\\mathbf x}, {\\mathbf y})\\text{.}$</p></article><article class=\"proof\" id=\"proof-47\"><h6 class=\"heading\"><span class=\"type\">Proof</span></h6><p id=\"p-1236\">Suppose that ${\\mathbf x}$ and ${\\mathbf y}$ are binary $n$-tuples. Then the distance between ${\\mathbf x}$ and ${\\mathbf y}$ is exactly the number of places in which ${\\mathbf x}$ and ${\\mathbf y}$ differ. But ${\\mathbf x}$ and ${\\mathbf y}$ differ in a particular coordinate exactly when the sum in the coordinate is 1, since</p><div class=\"displaymath\">\n\\begin{align*}\n1 + 1 & = 0\\\\\n0 + 0 & = 0\\\\\n1 + 0 & = 1\\\\\n0 + 1 & = 1.\n\\end{align*}\n</div><p>Consequently, the weight of the sum must be the distance between the two codewords.</p></article></div>", "_____no_output_____" ], [ "<div class=\"mathbook-content\"><article class=\"theorem-like\" id=\"theorem-33\"><h6 class=\"heading\"><span class=\"type\">Theorem</span><span class=\"codenumber\">8.18</span></h6><p id=\"p-1237\">Let $d_{\\min}$ be the minimum distance for a group code $C\\text{.}$ Then $d_{\\min}$ is the minimum of all the nonzero weights of the nonzero codewords in $C\\text{.}$ That is,</p><div class=\"displaymath\">\n\\begin{equation*}\nd_{\\min} = \\min\\{ w({\\mathbf x}) : { {\\mathbf x} \\neq {\\mathbf 0} } \\}.\n\\end{equation*}\n</div></article><article class=\"proof\" id=\"proof-48\"><h6 class=\"heading\"><span class=\"type\">Proof</span></h6><p id=\"p-1238\">Observe that</p><div class=\"displaymath\">\n\\begin{align*}\nd_{\\min} & = \\min \\{ d({\\mathbf x},{\\mathbf y}) : {\\mathbf x}\\neq{\\mathbf y} \\}\\\\\n&= \\min \\{ d({\\mathbf x},{\\mathbf y}) : {\\mathbf x}+{\\mathbf y} \\neq {\\mathbf 0} \\}\\\\\n&= \\min\\{ w({\\mathbf x} + {\\mathbf y}) : {\\mathbf x}+{\\mathbf y}\\neq {\\mathbf 0} \\}\\\\\n& = \\min\\{ w({\\mathbf z}) : {\\mathbf z} \\neq {\\mathbf 0} \\}.\n\\end{align*}\n</div></article></div>", "_____no_output_____" ], [ "<div class=\"mathbook-content\"><h3 class=\"heading hide-type\" alt=\"Subsection Linear Codes\"><span class=\"type\">Subsection</span><span class=\"codenumber\" /><span class=\"title\">Linear Codes</span></h3><a href=\"section-linear-codes.ipynb#algcodes-subsection-linear-codes\" class=\"permalink\">¶</a></div>", "_____no_output_____" ], [ "<div class=\"mathbook-content\"><p id=\"p-1239\">From Example <a href=\"section-linear-codes.ipynb#example-algcodes-weights\" class=\"xref\" alt=\"Example 8.16 \" title=\"Example 8.16 \">8.16</a>, it is now easy to check that the minimum nonzero weight is 3; hence, the code does indeed detect and correct all single errors. We have now reduced the problem of finding “good” codes to that of generating group codes. One easy way to generate group codes is to employ a bit of matrix theory.</p></div>", "_____no_output_____" ], [ "<div class=\"mathbook-content\"><p id=\"p-1240\">Define the <dfn class=\"terminology\">inner product</dfn> of two binary $n$-tuples to be</p><div class=\"displaymath\">\n\\begin{equation*}\n{\\mathbf x} \\cdot {\\mathbf y} = x_1 y_1 + \\cdots + x_n y_n,\n\\end{equation*}\n</div><p>where ${\\mathbf x} = (x_1, x_2, \\ldots, x_n)^{\\rm t}$ and ${\\mathbf y} = (y_1, y_2, \\ldots, y_n)^{\\rm t}$ are column vectors.<span class=\"footnote\"><a knowl=\"\" class=\"id-ref\" refid=\"hk-fn-4\" id=\"fn-4\">⁴</a></span><span id=\"hk-fn-4\" class=\"hidden-content tex2jax_ignore\"><span class=\"footnote\">Since we will be working with matrices, we will write binary $n$-tuples as column vectors for the remainder of this chapter.</span></span> For example, if ${\\mathbf x} = (011001)^{\\rm t}$ and ${\\mathbf y} = (110101)^{\\rm t}\\text{,}$ then ${\\mathbf x} \\cdot {\\mathbf y} = 0\\text{.}$ We can also look at an inner product as the product of a row matrix with a column matrix; that is,</p><div class=\"displaymath\">\n\\begin{align*}\n{\\mathbf x} \\cdot {\\mathbf y} & = {\\mathbf x}^{\\rm t} {\\mathbf y}\\\\\n& =\n\\begin{pmatrix}\nx_1 & x_2 & \\cdots & x_n\n\\end{pmatrix}\n\\begin{pmatrix}\ny_1 \\\\ y_2 \\\\ \\vdots \\\\ y_n\n\\end{pmatrix}\\\\\n& = x_{1}y_{1} + x_{2}y_{2} + \\cdots + x_{n}y_{n}.\n\\end{align*}\n</div></div>", "_____no_output_____" ], [ "<div class=\"mathbook-content\"><article class=\"example-like\" id=\"example-algcodes-matrixcodes\"><h6 class=\"heading\"><span class=\"type\">Example</span><span class=\"codenumber\">8.19</span></h6><p id=\"p-1241\">Suppose that the words to be encoded consist of all binary 3-tuples and that our encoding scheme is even-parity. To encode an arbitrary 3-tuple, we add a fourth bit to obtain an even number of 1s. Notice that an arbitrary $n$-tuple ${\\mathbf x} = (x_1, x_2, \\ldots, x_n)^{\\rm t}$ has an even number of 1s exactly when $x_1 + x_2 + \\cdots + x_n = 0\\text{;}$ hence, a 4-tuple ${\\mathbf x} = (x_1, x_2, x_3, x_4)^{\\rm t}$ has an even number of 1s if $x_1+ x_2+ x_3+ x_4 = 0\\text{,}$ or</p><div class=\"displaymath\">\n\\begin{equation*}\n{\\mathbf x} \\cdot {\\mathbf 1} = {\\mathbf x}^{\\rm t} {\\mathbf 1} =\n\\begin{pmatrix}\nx_1 & x_2 & x_3 & x_4\n\\end{pmatrix}\n\\begin{pmatrix}\n1 \\\\ 1 \\\\ 1 \\\\ 1\n\\end{pmatrix} = 0.\n\\end{equation*}\n</div><p>This example leads us to hope that there is a connection between matrices and coding theory.</p></article></div>", "_____no_output_____" ], [ "<div class=\"mathbook-content\"><p id=\"p-1242\">Let ${\\mathbb M}_{m \\times n}({\\mathbb Z}_2)$ denote the set of all $m \\times n$ matrices with entries in ${\\mathbb Z}_2\\text{.}$ We do matrix operations as usual except that all our addition and multiplication operations occur in ${\\mathbb Z}_2\\text{.}$ Define the <dfn class=\"terminology\">null space</dfn> of a matrix $H \\in {\\mathbb M}_{m \\times n}({\\mathbb Z}_2)$ to be the set of all binary $n$-tuples ${\\mathbf x}$ such that $H{\\mathbf x} = {\\mathbf 0}\\text{.}$ We denote the null space of a matrix $H$ by $\\Null(H)\\text{.}$ </p></div>", "_____no_output_____" ], [ "<div class=\"mathbook-content\"><article class=\"example-like\" id=\"example-algcodes-group-code\"><h6 class=\"heading\"><span class=\"type\">Example</span><span class=\"codenumber\">8.20</span></h6><p id=\"p-1243\">Suppose that</p><div class=\"displaymath\">\n\\begin{equation*}\nH =\n\\begin{pmatrix}\n0 & 1 & 0 & 1 & 0 \\\\\n1 & 1 & 1 & 1 & 0 \\\\\n0 & 0 & 1 & 1 & 1\n\\end{pmatrix}.\n\\end{equation*}\n</div><p>For a 5-tuple ${\\mathbf x} = (x_1, x_2, x_3, x_4, x_5)^{\\rm t}$ to be in the null space of $H\\text{,}$ $H{\\mathbf x} = {\\mathbf 0}\\text{.}$ Equivalently, the following system of equations must be satisfied:</p><div class=\"displaymath\">\n\\begin{align*}\nx_2 + x_4 & = 0\\\\\nx_1 + x_2 + x_3 + x_4 & = 0\\\\\nx_3 + x_4 + x_5 & = 0.\n\\end{align*}\n</div><p>The set of binary 5-tuples satisfying these equations is</p><div class=\"displaymath\">\n\\begin{equation*}\n(00000) \\qquad (11110) \\qquad (10101) \\qquad (01011).\n\\end{equation*}\n</div><p>This code is easily determined to be a group code.</p></article></div>", "_____no_output_____" ], [ "<div class=\"mathbook-content\"><article class=\"theorem-like\" id=\"theorem-34\"><h6 class=\"heading\"><span class=\"type\">Theorem</span><span class=\"codenumber\">8.21</span></h6><p id=\"p-1244\">Let $H$ be in ${\\mathbb M}_{m \\times n}({\\mathbb Z}_2)\\text{.}$ Then the null space of $H$ is a group code.</p></article><article class=\"proof\" id=\"proof-49\"><h6 class=\"heading\"><span class=\"type\">Proof</span></h6><p id=\"p-1245\">Since each element of ${\\mathbb Z}_2^n$ is its own inverse, the only thing that really needs to be checked here is closure. Let ${\\mathbf x}, {\\mathbf y} \\in {\\rm Null}(H)$ for some matrix $H$ in ${\\mathbb M}_{m \\times n}({\\mathbb Z}_2)\\text{.}$ Then $H{\\mathbf x} = {\\mathbf 0}$ and $H{\\mathbf y} = {\\mathbf 0}\\text{.}$ So</p><div class=\"displaymath\">\n\\begin{equation*}\nH({\\mathbf x}+{\\mathbf y}) = H{\\mathbf x} + H{\\mathbf y} = {\\mathbf 0} + {\\mathbf 0} = {\\mathbf 0}.\n\\end{equation*}\n</div><p>Hence, ${\\mathbf x} + {\\mathbf y}$ is in the null space of $H$ and therefore must be a codeword.</p></article></div>", "_____no_output_____" ], [ "<div class=\"mathbook-content\"><p id=\"p-1246\">A code is a <dfn class=\"terminology\">linear code</dfn> if it is determined by the null space of some matrix $H \\in {\\mathbb M}_{m \\times n}({\\mathbb Z}_2)\\text{.}$</p></div>", "_____no_output_____" ], [ "<div class=\"mathbook-content\"><article class=\"example-like\" id=\"example-algcodes-linear-code\"><h6 class=\"heading\"><span class=\"type\">Example</span><span class=\"codenumber\">8.22</span></h6><p id=\"p-1247\">Let $C$ be the code given by the matrix</p><div class=\"displaymath\">\n\\begin{equation*}\nH =\n\\begin{pmatrix}\n0 & 0 & 0 & 1 & 1 & 1 \\\\\n0 & 1 & 1 & 0 & 1 & 1 \\\\\n1 & 0 & 1 & 0 & 0 & 1\n\\end{pmatrix}.\n\\end{equation*}\n</div><p>Suppose that the 6-tuple ${\\mathbf x} = (010011)^{\\rm t}$ is received. It is a simple matter of matrix multiplication to determine whether or not ${\\mathbf x}$ is a codeword. Since</p><div class=\"displaymath\">\n\\begin{equation*}\nH{\\mathbf x} =\n\\begin{pmatrix} \n0 \\\\ 1 \\\\ 1\n\\end{pmatrix},\n\\end{equation*}\n</div><p>the received word is not a codeword. We must either attempt to correct the word or request that it be transmitted again.</p></article></div>", "_____no_output_____" ] ] ]

[ "code", "markdown" ]

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

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import tensorflow as tf\nfrom tensorflow.keras import datasets,layers,models\nimport matplotlib.pyplot as plt\nimport numpy as np", "_____no_output_____" ], [ "(X_train, y_train), (X_test, y_test) = datasets.cifar10.load_data()\nX_train.shape", "_____no_output_____" ], [ "X_test.shape", "_____no_output_____" ], [ "plt.imshow(X_train[0])", "_____no_output_____" ], [ "classes = ['airplane','automobile','bird','cat','deer','dog','frog','horse','ship','truck']", "_____no_output_____" ], [ "y_train = y_train.reshape(-1,)\nclasses[y_train[0]]", "_____no_output_____" ], [ "def plot_sample(X,y,index):\n plt.figure(figsize=(15,2))\n plt.imshow(X[index])\n plt.xlabel(classes[y[index]])", "_____no_output_____" ], [ "plot_sample(X_train,y_train,0)", "_____no_output_____" ], [ "plot_sample(X_train,y_train,4)", "_____no_output_____" ], [ "X_train = X_train/255\nX_test = X_test/255", "_____no_output_____" ], [ "#model building\nmodel = models.Sequential([\n layers.Flatten(input_shape=(32,32,3)),\n layers.Dense(3000,activation='relu'),\n layers.Dense(1000,activation='relu'),\n layers.Dense(10,activation='sigmoid'),\n])\n\nmodel.compile(\n optimizer='SGD',\n loss='sparse_categorical_crossentropy', \n metrics=['accuracy']\n)\n\nmodel.fit(X_train,y_train,epochs=5)", "Epoch 1/5\n1563/1563 [==============================] - 97s 62ms/step - loss: 1.8602 - accuracy: 0.3331\nEpoch 2/5\n1563/1563 [==============================] - 96s 61ms/step - loss: 1.6575 - accuracy: 0.41310s -\nEpoch 3/5\n1563/1563 [==============================] - 96s 62ms/step - loss: 1.5689 - accuracy: 0.4462\nEpoch 4/5\n1563/1563 [==============================] - 86s 55ms/step - loss: 1.5080 - accuracy: 0.4688\nEpoch 5/5\n1563/1563 [==============================] - 86s 55ms/step - loss: 1.4557 - accuracy: 0.4862\n" ], [ "tf.__version__", "_____no_output_____" ], [ "from sklearn.metrics import confusion_matrix, classification_report\nimport numpy as np\ny_pred = model.predict(X_test)\ny_pred_classes = [np.argmax(element) for element in y_pred]\n\nprint('Classification_report: \\n',classification_report(y_test,y_pred_classes))", "Classification_report: \n precision recall f1-score support\n\n 0 0.64 0.31 0.41 1000\n 1 0.70 0.49 0.58 1000\n 2 0.23 0.70 0.34 1000\n 3 0.32 0.41 0.36 1000\n 4 0.47 0.19 0.27 1000\n 5 0.44 0.25 0.32 1000\n 6 0.47 0.54 0.51 1000\n 7 0.69 0.32 0.44 1000\n 8 0.52 0.67 0.58 1000\n 9 0.67 0.39 0.49 1000\n\n accuracy 0.43 10000\n macro avg 0.51 0.43 0.43 10000\nweighted avg 0.51 0.43 0.43 10000\n\n" ], [ "cnn = models.Sequential([\n #cnn\n layers.Conv2D(filters=32,kernel_size=(3,3),activation='relu',input_shape=(32,32,3)),\n layers.MaxPool2D((2,2)),\n layers.Conv2D(filters=64,kernel_size=(3,3),activation='relu'),\n layers.MaxPool2D((2,2)),\n \n #dense\n layers.Flatten(),\n layers.Dense(64,activation='relu'),\n layers.Dense(10,activation='softmax')\n])\n\ncnn.compile(\n optimizer='SGD',\n loss='sparse_categorical_crossentropy',\n metrics=['accuracy']\n)\n\ncnn.fit(X_train,y_train,epochs=10)", "Epoch 1/10\n1563/1563 [==============================] - 46s 30ms/step - loss: 2.0075 - accuracy: 0.27360s - loss: 2.009\nEpoch 2/10\n1563/1563 [==============================] - 41s 26ms/step - loss: 1.6264 - accuracy: 0.4201\nEpoch 3/10\n1563/1563 [==============================] - 48s 31ms/step - loss: 1.4375 - accuracy: 0.4883\nEpoch 4/10\n1563/1563 [==============================] - 47s 30ms/step - loss: 1.3361 - accuracy: 0.5303\nEpoch 5/10\n1563/1563 [==============================] - 40s 26ms/step - loss: 1.2569 - accuracy: 0.5582\nEpoch 6/10\n1563/1563 [==============================] - 43s 27ms/step - loss: 1.1882 - accuracy: 0.5832\nEpoch 7/10\n1563/1563 [==============================] - 40s 26ms/step - loss: 1.1306 - accuracy: 0.6034\nEpoch 8/10\n1563/1563 [==============================] - 42s 27ms/step - loss: 1.0750 - accuracy: 0.6270\nEpoch 9/10\n1563/1563 [==============================] - 41s 26ms/step - loss: 1.0297 - accuracy: 0.6418\nEpoch 10/10\n1563/1563 [==============================] - 41s 26ms/step - loss: 0.9862 - accuracy: 0.6572\n" ], [ "cnn.evaluate(X_test,y_test)", "313/313 [==============================] - 2s 7ms/step - loss: 1.0495 - accuracy: 0.6361\n" ], [ "y_pred = cnn.predict(X_test)\ny_test = y_test.reshape(-1,)\ny_test", "_____no_output_____" ], [ "y_pred_classes = [np.argmax(element) for element in y_pred]\ny_pred_classes", "_____no_output_____" ], [ "plot_sample(X_test,y_test,2)", "_____no_output_____" ], [ "classes[y_pred_classes[2]]", "_____no_output_____" ], [ "print(\"Classification report: \\n\", classification_report(y_test,y_pred_classes))", "Classification report: \n precision recall f1-score support\n\n 0 0.70 0.70 0.70 1000\n 1 0.78 0.66 0.72 1000\n 2 0.58 0.45 0.50 1000\n 3 0.59 0.28 0.38 1000\n 4 0.55 0.62 0.58 1000\n 5 0.55 0.59 0.57 1000\n 6 0.62 0.83 0.71 1000\n 7 0.63 0.77 0.70 1000\n 8 0.82 0.68 0.75 1000\n 9 0.59 0.79 0.68 1000\n\n accuracy 0.64 10000\n macro avg 0.64 0.64 0.63 10000\nweighted avg 0.64 0.64 0.63 10000\n\n" ] ], [ [ "", "_____no_output_____" ] ], [ [ "(X_train,y_train),(X_test,y_test) = datasets.mnist.load_data()\nX_train.shape", "_____no_output_____" ], [ "X_train = X_train.reshape(60000,28,28,1)\nX_train.shape", "_____no_output_____" ], [ "X_train.shape", "_____no_output_____" ], [ "plt.imshow(X_train[0])", "_____no_output_____" ], [ "X_train = X_train/255\nX_test = X_test/255", "_____no_output_____" ], [ "mnist_cnn = models.Sequential([\n layers.Conv2D(filters=10, kernel_size=(5,5), activation='relu',input_shape=(28,28,1)),\n layers.MaxPooling2D(2,2),\n# layers.Conv2D(filters=5, kernel_size=(3,3), activation='relu',input_shape=(28,28,1)),\n# layers.MaxPooling2D(2,2),\n layers.Flatten(),\n layers.Dense(50,activation='relu'),\n layers.Dense(10,activation='softmax')\n])\n\nmnist_cnn.compile(\n optimizer='adam',\n loss='sparse_categorical_crossentropy', #categories 1,2,3... sparse because output is integer\n metrics=['accuracy']\n)\n\nmnist_cnn.fit(X_train,y_train,epochs=10)", "Epoch 1/10\n1875/1875 [==============================] - 12s 7ms/step - loss: 0.1976 - accuracy: 0.9426\nEpoch 2/10\n1875/1875 [==============================] - 13s 7ms/step - loss: 0.0706 - accuracy: 0.9786 0s - loss: 0.0707 - \nEpoch 3/10\n1875/1875 [==============================] - 12s 7ms/step - loss: 0.0491 - accuracy: 0.9850\nEpoch 4/10\n1875/1875 [==============================] - 13s 7ms/step - loss: 0.0388 - accuracy: 0.9878\nEpoch 5/10\n1875/1875 [==============================] - 13s 7ms/step - loss: 0.0303 - accuracy: 0.9903\nEpoch 6/10\n1875/1875 [==============================] - 13s 7ms/step - loss: 0.0245 - accuracy: 0.9920\nEpoch 7/10\n1875/1875 [==============================] - 13s 7ms/step - loss: 0.0192 - accuracy: 0.9935\nEpoch 8/10\n1875/1875 [==============================] - 13s 7ms/step - loss: 0.0157 - accuracy: 0.9948\nEpoch 9/10\n1875/1875 [==============================] - 13s 7ms/step - loss: 0.0126 - accuracy: 0.9960\nEpoch 10/10\n1875/1875 [==============================] - 13s 7ms/step - loss: 0.0113 - accuracy: 0.9961\n" ], [ "X_test.shape", "_____no_output_____" ], [ "X_test = X_test.reshape(10000,28,28,1)", "313/313 [==============================] - 1s 3ms/step - loss: 0.0560 - accuracy: 0.9843\n" ], [ "mnist_cnn.evaluate(X_test,y_test)", "313/313 [==============================] - 1s 3ms/step - loss: 0.0557 - accuracy: 0.9843\n" ], [ "y_pred = mnist_cnn.predict(X_test)\ny_pred_classes = [np.argmax(element) for element in y_pred]\ncm = confusion_matrix(y_test,y_pred_classes)\ncm", "_____no_output_____" ], [ "import seaborn as sn\nplt.figure(figsize=(10,7))\nsn.heatmap(cm,annot=True,fmt='d')\nplt.xlabel('Predicted')\nplt.ylabel('Truth')", "_____no_output_____" ] ] ]

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

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

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ [ [ "# Novel-taxa classification evaluation\n\nThe following notebooks describe the evaluation of taxonomy classifiers using \"novel taxa\" data sets. Novel-taxa analysis is a form of cross-validated taxonomic classification, wherein random unique sequences are sampled from the reference database as a test set; all sequences sharing taxonomic affiliation at a given taxonomic level are removed from the reference database (training set); and taxonomy is assigned to the query sequences at the given taxonomic level. Thus, this test interrogates the behavior of a taxonomy classifier when challenged with \"novel\" sequences that are not represented by close matches within the reference sequence database. Such an analysis is performed to assess the degree to which \"overassignment\" occurs for sequences that are not represented in a reference database.\n\nAt each level ``L``, the unique taxonomic clades are randomly sampled and used as ``QUERY`` sequences. All sequences that match that taxonomic annotation at ``L`` are excluded from ``REF``. Hence, species-level ``QUERY`` assignment asks how accurate assignment is to an \"unknown\" species that is not represented in the ``REF``, though other species in the same genus are. Genus-level ``QUERY`` assignment asks how accurate assignment is to an \"unknown\" genus that is not represented in the ``REF``, though other genera in the same family are, *et cetera*.\n\nThe steps involved in preparing and executing novel-taxa analysis are described in a series of notebooks:\n\n1) **[Novel taxa dataset generation](./dataset-generation.ipynb)** only needs to be performed once for a given reference database. Only run this notebook if you wish to make novel taxa datasets from a different reference database, or alter the parameters used to make the novel taxa datasets. The default included in Tax-Credit is Greengenes 13\\_8 release, amplified *in silico* with primers 515f and 806r, and trimmed to 250 nt from the 5' end.\n\n2) **[Taxonomic classification](./taxonomy-assignment.ipynb)** of novel taxa sequences is performed using the datasets generated in *step 1*. This template currently describes classification using QIIME 1 classifiers and can be used as a template for classifiers that are called via command line interface. Python-based classifiers can be used following the example of q2-feature-classifier.\n\n3) **[Classifier evaluation](./evaluate-classification.ipynb)** is performed based on taxonomic classifications generated by each classifier used in *step 2*. \n\n\n## Definitions\nThe **[dataset generation](./dataset-generation.ipynb)** notebook uses a few novel definitions. The following provides some explanation of the definitions used in that notebook.\n\n* ``source`` = original reference database sequences and taxonomy.\n* ``QUERY`` = 'novel' query sequences and taxonomies randomly drawn from ``source``. \n* ``REF`` = ``source`` - ``novel`` taxa, used for taxonomy assignment.\n* ``L`` = taxonomic level being tested\n * 0 = kingdom, 1 = phylum, 2 = class, 3 = order, 4 = family, 5 = genus, 6 = species\n* ``branching`` = describes a taxon at level ``L`` that \"branches\" into two or more lineages at ``L + 1``. \n * A \"branched\" taxon, then, describes these lineages. E.g., in the example below Lactobacillaceae, Lactobacillus, and Pediococcus branch, while Paralactobacillus is unbranching. The Lactobacillus and Pediococcus species are \"branched\". Paralactobacillus selangorensis is \"unbranched\"\n * The novel taxa analysis only uses \"branching\" taxa, such that for each ``QUERY`` at level ``L``, ``REF`` must contain one or more taxa that share the same clade at level ``L - 1``.\n\n```\nLactobacillaceae\n └── Lactobacillus\n │ ├── Lactobacillus brevis\n │ └── Lactobacillus sanfranciscensis\n ├── Pediococcus\n │ ├── Pediococcus damnosus\n │ └── Pediococcus claussenii\n └── Paralactobacillus\n └── Paralactobacillus selangorensis\n```\n", "_____no_output_____" ] ] ]

[ "markdown" ]

[ [ "markdown" ] ]

[ "CC0-1.0" ]

[ "CC0-1.0" ]

[ "CC0-1.0" ]

[ [ [ "# Importing necessary libraries\n#===============================================================================\nimport matplotlib as mpl\nmpl.use('qt5agg')\nmpl.interactive(True)\nimport matplotlib.pyplot as plt\nimport numpy as np\nimport pandas as pd\nimport seaborn as sbn\nsbn.set()\nfrom scipy import stats\nimport h5py\nfrom os.path import dirname\nfrom pathlib import Path\nimport sys\nimport mmodel_reversals as mm", "_____no_output_____" ], [ "# Setting paths\n#===============================================================================\nROOTPATH = Path().cwd()\n(ROOTPATH / 'Export').mkdir(parents=True, exist_ok=True)", "_____no_output_____" ], [ "# Function to compute complex-valued OLS\n#===============================================================================\ndef complexGLM(pred, crit):\n '''\n Compute regression weights for predicting the criterion variable using predictor arrays\n In -> pred = predictor array, crit = criterion vector\n Out -> coefs = regression coefficients/weights\n '''\n pred = np.array(pred)\n crit = np.array(crit)\n if len(crit.shape) < 2:\n crit = crit.reshape(-1, 1)\n if pred.dtype is not np.dtype('complex'):\n pred = np.exp(pred * 1j)\n if crit.dtype is not np.dtype('complex'):\n crit = np.exp(crit * 1j)\n a, b = [crit.shape[0], pred.shape[0]]\n if crit.shape[0] != pred.shape[0]:\n raise ValueError('The two arrays are of incompatible shape, {} and {}'.format(a, b))\n coefs = np.asmatrix(np.asmatrix(pred).H * np.asmatrix(pred)).I * (np.asmatrix(pred).H * np.asmatrix(crit))\n return coefs", "_____no_output_____" ], [ "# Setting simulation parameters\n#===============================================================================\nnp.random.seed(0)\ntrlN = 1000\nrunN = 10000\nsimK = np.sort([.1, 2.5, 1., 5., 10.])", "_____no_output_____" ], [ "# Simulate independently sampled motion directions\n#===============================================================================\npresDirs_ind = np.angle(\n np.exp(\n np.random.uniform(\n 0, 2 * np.pi, \n size = [runN, trlN, 6]\n ) * 1j\n )\n)\n\npercDirs_ind = np.concatenate([\n np.angle(\n np.exp(\n np.array(\n [\n np.random.vonmises(\n presDirs_ind, K\n )\n for K in simK\n ]\n ) * 1j\n )\n ),\n # no noise condition, K = inf\n presDirs_ind[None]\n])\n# saving data for independently sampled directions\nwith h5py.File(ROOTPATH / 'Export' / 'simData.hdf', 'a') as f:\n f.create_dataset(\n name = 'presDirs_ind', \n data = presDirs_ind, \n compression = 9\n )\nwith h5py.File(ROOTPATH / 'Export' / 'simData.hdf', 'a') as f:\n f.create_dataset(\n name = 'percDirs_ind', \n data = percDirs_ind, \n compression = 9\n )\npresDirs_ind = None\npercDirs_ind = None", "_____no_output_____" ], [ "# Simulate dependently sampled motion direction\n#===============================================================================\nfrstTar, frstFoil = np.random.choice(\n np.arange(0, 360), \n size = [2, runN, trlN]\n)\nfrstDis, scndTar = (\n frstTar[None] \n # random direction (CW/CCW)\n + np.random.choice(\n [-1, 1],\n size = [2, runN, trlN]\n ) \n # random angular offset\n * np.random.choice(\n np.arange(30, 151),\n size = [2, runN, trlN]\n )\n)\nscndDis, scndFoil = (\n np.stack(\n [scndTar, frstFoil]\n )\n # random direction (CW/CCW)\n + np.random.choice(\n [-1, 1],\n size = [2, runN, trlN]\n ) \n # random angular offset\n * np.random.choice(\n np.arange(30, 151),\n size = [2, runN, trlN]\n )\n)\npresDirs_dep = np.angle(\n np.exp(\n np.deg2rad(np.stack(\n [frstTar, scndTar, frstDis, scndDis, frstFoil, scndFoil],\n axis = -1\n )) * 1j\n )\n)\n\npercDirs_dep = np.concatenate([\n np.angle(\n np.exp(\n np.array(\n [\n np.random.vonmises(\n presDirs_dep, K\n )\n for K in simK\n ]\n ) * 1j\n )\n ),\n # no noise condition, K = inf\n presDirs_dep[None]\n])\n\n# saving data for dependently sampled directions\nwith h5py.File(ROOTPATH / 'Export' / 'simData.hdf', 'a') as f:\n f.create_dataset(\n name = 'presDirs_dep', \n data = presDirs_dep, \n compression = 9\n )\nwith h5py.File(ROOTPATH / 'Export' / 'simData.hdf', 'a') as f:\n f.create_dataset(\n name = 'percDirs_dep', \n data = percDirs_dep, \n compression = 9\n )\npresDirs_dep = None\npercDirs_dep = None", "_____no_output_____" ], [ "# Simulate complex-valued regression weights\n#===============================================================================\nsimCoefAbs = np.random.uniform(size = [runN, 6])\n# the angles of weigthing coeficients\nsimCoefAng = np.random.uniform(\n 0, 2 * np.pi,\n size = [runN, 6]\n)\nwith h5py.File(ROOTPATH / 'Export' / 'simData.hdf', 'a') as f:\n f.create_dataset(\n name = 'coefsAbs', \n data = simCoefAbs, \n compression = 9\n )\n f.create_dataset(\n name = 'coefsAng', \n data = simCoefAng, \n compression = 9\n )\nsimCoefAbs = None\nsimCoefAng = None", "_____no_output_____" ], [ "# Run complex-valued OLS for different simulation conditions\n#===============================================================================\nfor cond in ['ind', 'dep', 'dep_ss']:\n # there are three conditions:\n # ind: independently sampled motion\n # dep: dependently sampled motion\n # dep_ss: dependently sampled motion, 100 trials per run\n print('Analysing {} simulation condition'.format(cond.upper()))\n ssize = None\n cond_raw = cond\n if 'ss' in cond.split('_'):\n cond, ssize = cond.split('_')\n with h5py.File(ROOTPATH / 'Export' / 'simData.hdf', 'r') as f:\n presDirs = f['presDirs_{}'.format(cond)][:]\n percDirs = f['percDirs_{}'.format(cond)][:]\n coefsAbs = f['coefsAbs'][:]\n coefsAng = f['coefsAng'][:]\n if ssize:\n presDirs = presDirs[:, :100]\n percDirs = percDirs[:, :, :100]\n\n # running complex-values OLS for different simulated weight angles\n for idx_simAngle, simAngle in enumerate(['null', 'real']):\n # two analyses are run\n # null: the angles of the simulated complex-valued regression weights are zero\n # real: the angles are are randomly sampled \n simCoefs = (\n np.exp(\n [0, 1][idx_simAngle] * coefsAng * 1j\n ) * coefsAbs\n ) \n # %% simulating response on the basis of perceived directions and simulated\n respDirs = np.array([\n np.angle(\n np.sum(\n simCoefs[:, None] \n * np.exp(simKappa * 1j), \n -1))\n for simKappa in percDirs\n ])\n # weighting coefficients\n coefs = np.array(\n [\n [\n complexGLM(presDirs[idxRun], run)\n for idxRun, run in enumerate(simKappa)\n ]\n for simKappa in respDirs\n ] \n ).squeeze()\n print('Finished complex OLS')\n # %% goodness of fit\n predDirs = np.array([\n np.angle(\n np.sum(\n simKappa[:, None, :] \n * np.exp(presDirs * 1j), -1\n )\n )\n for simKappa in coefs\n ])\n GoF = np.array([\n np.angle(\n np.exp(respDirs[simKappa] * 1j)\n / np.exp(predDirs[simKappa] * 1j)\n )\n for simKappa in range(coefs.shape[0])\n ])\n # saving data\n with h5py.File(ROOTPATH / 'Export' / 'simCoefs.hdf', 'a') as f:\n f.create_dataset(\n name = 'coefsAbsHat_{}_{}'.format(cond_raw,simAngle), \n data = np.abs(coefs), \n compression = 9\n )\n f.create_dataset(\n name = 'coefsAngHat_{}_{}'.format(cond_raw,simAngle), \n data = np.angle(coefs), \n compression = 9\n )\n f.create_dataset(\n name = 'GoF_{}_{}'.format(cond_raw,simAngle), \n data = GoF, \n compression = 9\n )", "_____no_output_____" ], [ "# Setting parameters for plotting supplementary figure 1\n#===============================================================================\n# two different plottings can be performed\n# first, the results for simulated complex-valued weights using real angles\n# second, the results for simulated weights using zero angles\n# here, only the real values are plotted.\n# N.B., the results for zero angles yields similart goodness-of-fit\n# N.B., the ability of the complex-valued OLS to recover the angles (not plotted)\n# is similar to its ability to recover the lengths, i.e., the decision weights .\nconds = [\n 'GoF_ind_real',\n 'GoF_dep_real',\n 'GoF_dep_ss_real'\n]\nwith h5py.File(ROOTPATH / 'Export' / 'simCoefs.hdf', 'r') as f:\n GoF = dict([(cond, f[cond][:]) for cond in conds])", "_____no_output_____" ], [ "# Plotting supplementary figure 1\n#===============================================================================\nsbn.set_style('ticks')\nSSIZE = 8\nMSIZE = 10\nLSIZE = 12\nparams = {'lines.linewidth' : 1.5,\n 'grid.linewidth' : 1,\n 'xtick.labelsize' : MSIZE,\n 'ytick.labelsize' : MSIZE,\n 'xtick.major.width' : 1,\n 'ytick.major.width' : 1,\n 'xtick.major.size' : 5,\n 'ytick.major.size' : 5,\n 'xtick.direction' : 'inout',\n 'ytick.direction' :'inout',\n 'axes.linewidth': 1,\n 'axes.labelsize' : MSIZE,\n 'axes.titlesize' : MSIZE,\n 'figure.titlesize' : LSIZE,\n 'font.size' : MSIZE,\n 'savefig.dpi': 300,\n 'font.sans-serif' : ['Calibri'],\n 'legend.fontsize' : MSIZE,\n 'hatch.linewidth' : .2}\nsbn.mpl.rcParams.update(params)\ncols = sbn.husl_palette(6, h = .15, s = .75, l = .5)\nsimK = np.sort([.1, 2.5, 1., 5., 10.])\nsimNoise = np.random.vonmises(0, simK[:, None], [5, 100000])\nfig = plt.figure(figsize = (8,2.8))\nax = fig.add_subplot(1, 4, 1)\nfor idx_noise, noise in enumerate(simNoise):\n sbn.kdeplot(\n noise,\n color = cols[idx_noise],\n alpha = .8,\n lw = 2,\n label = simK[idx_noise],\n ax = ax\n )\nax.axvline(0, color = cols[-1], alpha = .8, lw = 2, label = 'No noise')\nfor idx_cond, cond in enumerate(conds):\n ax = fig.add_subplot(1,4,2 + idx_cond)\n for idxK, err in enumerate(GoF[cond]):\n sbn.kdeplot(\n err.flatten(),\n color = cols[idxK],\n alpha = .8,\n lw = 2,\n label = '{}$\\degree$'.format(\n np.rad2deg(mm.cstd(err.flatten())).astype('int')\n ),\n ax = ax\n )\nfor idx_ax, ax in enumerate(fig.axes):\n title = '$\\kappa$'\n xlab = 'Perceptual noise'\n if idx_ax:\n title = '$\\sigma$'\n xlab = 'Prediction error'\n ax.legend(\n title = title, \n frameon = False,\n handlelength = 1,\n handletextpad = .5,\n markerfirst = False\n )\n ax.set_ylim(-0.05, 7)\n ax.set_xlim(-np.pi*1.1, np.pi*1.1)\n ax.set_xticks([-np.pi, 0, np.pi])\n ax.set_xticklabels(['-$\\pi$', '0', '$\\pi$'])\n ax.set_yticks([])\n ax.set_xlabel(xlab)\n ax.set_ylabel('Probability density')\n sbn.despine(ax = ax)\n ax.spines['bottom'].set_bounds(-np.pi, np.pi)\n ax.spines['left'].set_visible(False)\n if idx_ax:\n ax.yaxis.set_visible(False)\nplt.tight_layout(rect = (0, 0, 1, 1))\nfig.savefig(\n str(ROOTPATH / 'Export'/ 'GoodnessOfFit_All.png'), \n dpi = 600\n)\nplt.close(fig)", "_____no_output_____" ], [ "# Setting parameters for plotting supplementary figure 2\n#===============================================================================\nconds = [\n 'ind_real',\n 'dep_real',\n 'dep_ss_real'\n]\nwith h5py.File(ROOTPATH / 'Export' / 'simData.hdf', 'r') as f:\n coefsAbs = f['coefsAbs'][:]\ncols = sbn.husl_palette(6, h = .15, s = .75, l = .5)", "_____no_output_____" ], [ "# Plotting panels A-C of supplementary figure 2\n#===============================================================================\nfor idx_cond, cond in enumerate(conds):\n fig = plt.figure(figsize = (4,2.8))\n with h5py.File(ROOTPATH / 'Export' / 'simCoefs.hdf', 'r') as f:\n coefsAbsHat = f['_'.join(['coefsAbsHat', cond])][:]\n for idxK, weights in enumerate(coefsAbsHat):\n ax = fig.add_subplot(2, 3, idxK + 1)\n scatter = ax.plot(\n coefsAbs.flatten(), \n weights.flatten(), \n '.',\n mec = (.9,.9,.9),\n mfc = 'none',\n zorder = -10\n )\n line = ax.plot(\n np.array([0, 1]), np.array([0, 1]), \n 'k--',\n lw = 1,\n zorder = 0\n )\n bins = pd.qcut(coefsAbs.flatten(), 4).codes\n dataset = [weights.flatten()[bins == bin] for bin in np.unique(bins)]\n vlnplt = ax.violinplot(\n dataset, \n positions = [.125, .375, .625, .875],\n showextrema = False,\n showmedians = True,\n widths = .15,\n ) \n for i in vlnplt['bodies']:\n i.set_alpha(.8)\n i.set_facecolor(cols[idxK])\n i.set_lw(0)\n vlnplt['cmedians'].set_edgecolor('white')\n vlnplt['cmedians'].set_lw(.5)\n ax.text(\n .05, .95,\n (\n ['$\\kappa$ = {}'.format(k) for k in simK] \n + ['No noise']\n )[idxK],\n transform = ax.transAxes,\n va = 'top'\n )\n ax.set_xlabel('Simulated weights')\n ax.set_ylabel('Estimated weights')\n for idx_ax, ax in enumerate(fig.axes):\n ax.tick_params('both', direction = 'out')\n ax.set_xlim(-.1, 1.1)\n ax.set_ylim(-.1, 1.1)\n ax.spines['bottom'].set_bounds(0,1)\n ax.spines['left'].set_bounds(0,1)\n ax.spines['top'].set_visible(False)\n ax.spines['right'].set_visible(False)\n ax.set_xticks(np.linspace(0, 1, 3))\n ax.set_yticks(np.linspace(0, 1, 3))\n if idx_ax not in [0, 3]:\n ax.yaxis.set_visible(False)\n ax.spines['left'].set_visible(False)\n if idx_ax not in [3, 4, 5]:\n ax.xaxis.set_visible(False)\n ax.spines['bottom'].set_visible(False)\n plt.tight_layout(rect = (0, 0, 1, .975))\n label = [\n 'Independently sampled motion, 10$^3$ trials, 10$^4$ runs',\n 'Dependently sampled motion, 10$^3$ trials, 10$^4$ runs',\n 'Dependently sampled motion, 10$^2$ trials, 10$^4$ runs'\n ][idx_cond]\n fig.text(\n .5, 1, \n label,\n ha = 'center',\n va = 'top'\n )\n fig.savefig(\n str(\n ROOTPATH / \n 'Export' / \n 'WeightRecovery_{}.png'\n ).format([\n 'A', 'B', 'C'\n ][idx_cond]),\n dpi = 600\n )\n plt.close(fig)", "_____no_output_____" ], [ "# Plotting panel D of supplementary figure 2\n#===============================================================================\nfrom mpl_toolkits.axes_grid1 import ImageGrid\ncols = sbn.husl_palette(6, h = .15, s = .75, l = .5)\nfig = plt.figure(figsize = (4,2.8))\ngrid = ImageGrid(\n fig, 111, nrows_ncols = (2, 3), \n share_all = True, cbar_mode= 'single', aspect= True\n)\nfor idxK, weights in enumerate(coefsAbsHat):\n ax = grid[idxK]\n heatmap, xedges, yedges = np.histogram2d(\n np.array(list(map(\n stats.rankdata,\n coefsAbs\n ))).flatten(), \n np.array(list(map(\n stats.rankdata,\n weights\n ))).flatten(),\n bins = np.linspace(.5, 6.5, 7)\n )\n heatmap /= heatmap.sum()\n extent = [xedges[0], xedges[-1], yedges[0], yedges[-1]]\n im = ax.imshow(\n heatmap, \n extent = extent, origin = 'lower', \n vmin = 0, vmax = .15,\n cmap = 'viridis'\n )\n ax.text(\n .05, .95,\n (\n ['$\\kappa$ = {}'.format(k) for k in simK] \n + ['No noise']\n )[idxK],\n transform = ax.transAxes,\n va = 'top',\n color = 'white'\n )\ngrid.cbar_axes[0].colorbar(im)\ngrid.cbar_axes[0].set_ylim(0, .14)\ngrid.cbar_axes[0].set_yticks([.0, .05, .10, .15])\ngrid.cbar_axes[0].set_yticklabels(['0','5','10', '15'])\ngrid.cbar_axes[0].tick_params(direction = 'inout', length = 5)\ngrid[0].tick_params('both', direction = 'out', length = 5)\nfor idx_ax, ax in enumerate(grid):\n ax.tick_params('both', direction = 'inout', length = 5)\n ax.set_yticks(np.linspace(1,6,6))\n ax.set_xticks(np.linspace(1,6,6))\n if idx_ax not in [0, 3]:\n ax.yaxis.set_visible(False)\n if idx_ax < 3:\n ax.xaxis.set_visible(False)\nplt.tight_layout(rect = (.01, .01, .94, .99))\nfig.text(\n .5, .99, \n 'Dependently sampled motion, 10$^2$ trials, 10$^4$ runs',\n ha = 'center',\n va = 'top'\n)\nfig.text(\n .01, .5,\n 'Estimated weight rank',\n ha = 'left',\n va = 'center',\n rotation = 90\n)\nfig.text(\n .5, .01,\n 'Simulated weight rank',\n ha = 'center',\n va = 'bottom',\n)\nfig.text(\n .99, .5,\n 'Frequency [%]',\n ha = 'right',\n va = 'center',\n rotation = -90\n)\nfig.savefig(\n str(\n ROOTPATH /\n 'Export' /\n 'WeightRecovery_D.png'\n ), \n dpi = 600\n)\nplt.close(fig)", "_____no_output_____" ] ] ]

[ "code" ]

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

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "## EDA for Import and Export Trade Volumes", "_____no_output_____" ], [ "### Binational trade relationship between Mexico and the United States", "_____no_output_____" ] ], [ [ "#import key libraries\nimport pandas as pd\nimport matplotlib.pyplot as plt\n%matplotlib inline", "_____no_output_____" ] ], [ [ "## Dataset 1: General Imports from Mexico to the United States", "_____no_output_____" ] ], [ [ "imports = pd.read_csv(\"./data/usitc/total-imports-mx2us.csv\")", "_____no_output_____" ], [ "## data to be read includes the customs value of the import and the year\nimports.shape", "_____no_output_____" ], [ "imports.head()\n#note that the customs_value and the dollar_amount are the same just different data types", "_____no_output_____" ], [ "list(imports.columns)", "_____no_output_____" ], [ "imports['imports'].describe()", "_____no_output_____" ], [ "imports['dollar_amount'].describe()", "_____no_output_____" ], [ "imports['customs_value'].plot(kind=\"bar\")\n## confirming that the data is linear", "_____no_output_____" ], [ "plt.scatter(imports[\"year\"],imports['customs_value'],color=\"blue\")\nplt.title('Imports from Mexico to the US, Annual')\nplt.xlabel('year')\nplt.ylabel('customs value e11')\nplt.show()\n##amazing! Looks pretty linear to me", "_____no_output_____" ] ], [ [ "## Dataset #2 Exports from US to Mexico", "_____no_output_____" ] ], [ [ "exports = pd.read_csv(\"./data/usitc/total-exports-us2mx.csv\")", "_____no_output_____" ], [ "exports.shape", "_____no_output_____" ], [ "exports.head()", "_____no_output_____" ], [ "list(exports.columns)", "_____no_output_____" ], [ "exports['exports'].describe()", "_____no_output_____" ], [ "plt.scatter(exports[\"year\"],exports['exports'],color=\"green\")\nplt.title('Exports from US to Mexico, Annual')\nplt.xlabel('year')\nplt.ylabel('FAS Value e11')\nplt.show()\n##generally pretty linear", "_____no_output_____" ], [ "## Combining both exports and imports", "_____no_output_____" ], [ "##combine both vectors on one graph\nplt.plot(exports[\"year\"],exports['exports'],color=\"green\")\nplt.scatter(imports[\"year\"],imports['imports'],color=\"blue\")\nplt.title(\"Plotting imports and exports\")\nplt.xlabel(\"Year\")\nplt.ylabel(\"Value\")\nplt.legend()\nplt.show()", "_____no_output_____" ] ], [ [ "## Data preprocessing", "_____no_output_____" ] ], [ [ "# imports\nyear_var = list(imports['year'])\nprint(year_var)", "_____no_output_____" ], [ "dollar = list(imports[\"dollar_amount\"])\nprint(dollar)", "_____no_output_____" ], [ "def pre_process(year, dollar):\n print(\"[\",year,\",\",dollar,\"]\",\",\")", "_____no_output_____" ], [ "pre_process(1996, 2)", "_____no_output_____" ] ], [ [ "## Running descriptive statistics", "_____no_output_____" ] ], [ [ "# Pulling in descriptive statistics on IMPORTS\nfrom scipy import stats\nstats.describe(ytrain_pred)", "_____no_output_____" ], [ "imports['imports'].describe()", "_____no_output_____" ], [ "exports[\"exports\"].describe()", "_____no_output_____" ] ] ]

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

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

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "empty" ] ] ]

[ "empty" ]

[ [ "empty" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "## 1: Introduction To The Dataset", "_____no_output_____" ] ], [ [ "data = open('US_births_1994-2003_CDC_NCHS.csv','r').read().split('\\n')\ndata[:10]", "_____no_output_____" ] ], [ [ "## 2: Converting Data Into A List Of Lists", "_____no_output_____" ] ], [ [ "def read_csv(filename,header = False):\n final_list = []\n read_data = open(filename,'r').read().split('\\n')[1:]\n if header == True:\n read_data = open(filename,'r').read().split('\\n')[1:]\n else:\n read_data = open(filename,'r').read().split('\\n')\n for item in read_data:\n int_fields = []\n string_fields = item.split(',')\n for val in string_fields:\n int_fields.append(int(val))\n final_list.append(int_fields) \n return(final_list) \ncdc_list = read_csv('US_births_1994-2003_CDC_NCHS.csv',header = True)\ncdc_list[:10]", "_____no_output_____" ] ], [ [ "## 3: Calculating Number Of Births Each Month", "_____no_output_____" ] ], [ [ "def month_births(data):\n births_per_month = {}\n for item in data:\n if item[1] in births_per_month.keys():\n births_per_month[item[1]] += item[4]\n else:\n births_per_month[item[1]] = item[4]\n return(births_per_month)\ncdc_month_births = month_births(cdc_list) \ncdc_month_births", "_____no_output_____" ], [ "def dow_births(data):\n births_per_dow = {}\n for item in data:\n if item[3] in births_per_dow.keys():\n births_per_dow[item[3]] += item[4]\n else:\n births_per_dow[item[3]] = item[4]\n return(births_per_dow)\ncdc_day_births = dow_births(cdc_list) \ncdc_day_births", "_____no_output_____" ] ], [ [ "## 5: Creating A More General Function", "_____no_output_____" ] ], [ [ "def calc_counts(data,column):\n birth = {}\n for item in data:\n if item[column] in birth.keys():\n birth[item[column]] += item[4]\n else:\n birth[item[column]] = item[4]\n return(birth)\ncdc_year_births = calc_counts(cdc_list, 0)\ncdc_month_births = calc_counts(cdc_list, 1)\ncdc_dom_births = calc_counts(cdc_list, 2)\ncdc_dow_births = calc_counts(cdc_list, 3)", "_____no_output_____" ], [ "cdc_year_births", "_____no_output_____" ], [ "cdc_month_births", "_____no_output_____" ], [ "cdc_dom_births", "_____no_output_____" ], [ "cdc_dow_births", "_____no_output_____" ], [ "def min_max(dictionary):\n min_val = min(dictionary.items(), key=lambda k: k[1])\n max_val = max(dictionary.items(), key=lambda k: k[1])\n return(\"Minimum Value:%s Maximum Value:%s\"%(min_val,max_val))\nmin_max(cdc_dow_births)", "_____no_output_____" ] ] ]

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

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

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ [ [ "import matplotlib.pyplot as plt\nfrom mpl_toolkits.mplot3d import Axes3D\nfrom csv import reader\nimport numpy\n\niris = numpy.genfromtxt('data/iris.csv', delimiter=',', dtype=\"|U5\")\nX = iris[1:,:2].astype(numpy.float)\ny = numpy.unique(iris[1:,4], return_inverse=True)[1]\n\nplt.scatter(X[:, 0], X[:, 1], c=y)\nplt.xlim(3.5, 8.5)\nplt.ylim(1.5, 5)", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "from pandas import DataFrame\nimport matplotlib.pyplot as plt\nfrom sklearn.cluster import KMeans", "_____no_output_____" ], [ "\nData = {'x': [25,34,22,27,23,24,31,22,35,26,28,54,57,43,36,27,29,52,32,47,39,48,35,33,44,45,38,43,41,46],\n 'y': [79,51,53,78,99,92,73,57,69,75,51,32,40,77,53,36,35,58,59,50,25,20,14,12,20,5,29,27,8,7]\n }", "_____no_output_____" ], [ "df = DataFrame(Data,columns=['x','y'])", "_____no_output_____" ], [ "kmeans = KMeans(n_clusters=2).fit(df)\ncentroids = kmeans.cluster_centers_\nprint(centroids)", "[[31.6875 67.125 ]\n [41.35714286 22.14285714]]\n" ], [ "plt.scatter(df['x'], df['y'], c= kmeans.labels_.astype(float), s=50, alpha=0.6)\nplt.scatter(centroids[:, 0], centroids[:, 1], c='blue', s=50)\nplt.title('Kişi Sonuç Değerlendirmesi')\nplt.ylabel('Başarı Puanı')\nplt.xlabel('Yaş ')\nplt.show()", "_____no_output_____" ] ] ]

[ "code" ]

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

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import os\nimport pandas as pd\nimport matplotlib.pyplot as plt\nfrom Bio import SeqIO\nimport csv\nimport sqlite3\nimport time", "_____no_output_____" ], [ "con=sqlite3.connect(\"80kprotein_stats.db\")\ncur = con.cursor()\ndef get_protein_seq(pid):\n #cmd = \"SELECT * FROM proteins WHERE hashid='%s'\" % pid\n cmd = \"SELECT * FROM proteins WHERE pid = '%s'\" % pid \n #print(cmd)\n cur.execute(cmd)\n return cur.fetchone()[-1]", "_____no_output_____" ], [ "path_clu_rep = \"../clusters/clu_rep_stringent_final.csv\"", "_____no_output_____" ], [ "outfile = open(\"../clusters/INPUT/permissive/clu_perm_mmseqs_input.faa\", \"w\")\nwith open(path_clu_rep, 'r') as file:\n reader = csv.reader(file)\n next(reader)\n prev_rep = \"\"\n for row in reader:\n stringent_rep = row[0]\n if prev_rep != stringent_rep:\n print(\">\" + stringent_rep, file = outfile)\n print(get_protein_seq(stringent_rep), file = outfile)\n prev_rep = stringent_rep", "_____no_output_____" ], [ "con.commit()\ncon.close()", "_____no_output_____" ] ] ]

[ "code" ]

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

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "<a href=\"https://cognitiveclass.ai/\">\n <img src=\"https://s3-api.us-geo.objectstorage.softlayer.net/cf-courses-data/CognitiveClass/PY0101EN/Ad/CCLog.png\" width=\"200\" align=\"center\">\n</a>", "_____no_output_____" ], [ "<h1>Classes and Objects in Python</h1>", "_____no_output_____" ], [ "<p>\n <strong>Welcome!</strong> \n Objects in programming are like objects in real life. Like life, there are different classes of objects. In this notebook, we will create two classes called Circle and Rectangle. By the end of this notebook, you will have a better idea about :\n <ul>\n <li>what a class is</li>\n <li>what an attribute is</li>\n <li>what a method is</li>\n </ul>\n\n Don’t worry if you don’t get it the first time, as much of the terminology is confusing. Don’t forget to do the practice tests in the notebook.\n</p>", "_____no_output_____" ], [ "<div class=\"alert alert-block alert-info\" style=\"margin-top: 20px\">\n <a href=\"https://cocl.us/topNotebooksPython101Coursera\">\n <img src=\"https://s3-api.us-geo.objectstorage.softlayer.net/cf-courses-data/CognitiveClass/PY0101EN/Ad/TopAd.png\" width=\"750\" align=\"center\">\n </a>\n</div>", "_____no_output_____" ], [ "<h2>Table of Contents</h2>\n<div class=\"alert alert-block alert-info\" style=\"margin-top: 20px\">\n <ul>\n <li>\n <a href=\"#intro\">Introduction to Classes and Objects</a>\n <ul>\n <li><a href=\"create\">Creating a class</a></li>\n <li><a href=\"instance\">Instances of a Class: Objects and Attributes</a></li>\n <li><a href=\"method\">Methods</a></li>\n </ul>\n </li>\n <li><a href=\"creating\">Creating a class</a></li>\n <li><a href=\"circle\">Creating an instance of a class Circle</a></li>\n <li><a href=\"rect\">The Rectangle Class</a></li>\n </ul>\n <p>\n Estimated time needed: <strong>40 min</strong>\n </p>\n</div>\n\n<hr>", "_____no_output_____" ], [ "<h2 id=\"intro\">Introduction to Classes and Objects</h2>", "_____no_output_____" ], [ "<h3>Creating a Class</h3>", "_____no_output_____" ], [ "The first part of creating a class is giving it a name: In this notebook, we will create two classes, Circle and Rectangle. We need to determine all the data that make up that class, and we call that an attribute. Think about this step as creating a blue print that we will use to create objects. In figure 1 we see two classes, circle and rectangle. Each has their attributes, they are variables. The class circle has the attribute radius and color, while the rectangle has the attribute height and width. Let’s use the visual examples of these shapes before we get to the code, as this will help you get accustomed to the vocabulary.", "_____no_output_____" ], [ "<img src=\"https://s3-api.us-geo.objectstorage.softlayer.net/cf-courses-data/CognitiveClass/PY0101EN/Chapter%203/Images/ClassesClass.png\" width=\"500\" />", "_____no_output_____" ], [ "<i>Figure 1: Classes circle and rectangle, and each has their own attributes. The class circle has the attribute radius and colour, the rectangle has the attribute height and width.</i>\n", "_____no_output_____" ], [ "<h3 id=\"instance\">Instances of a Class: Objects and Attributes</h3>", "_____no_output_____" ], [ "An instance of an object is the realisation of a class, and in Figure 2 we see three instances of the class circle. We give each object a name: red circle, yellow circle and green circle. Each object has different attributes, so let's focus on the attribute of colour for each object.", "_____no_output_____" ], [ "<img src=\"https://s3-api.us-geo.objectstorage.softlayer.net/cf-courses-data/CognitiveClass/PY0101EN/Chapter%203/Images/ClassesObj.png\" width=\"500\" />", "_____no_output_____" ], [ "<i>Figure 2: Three instances of the class circle or three objects of type circle.</i>", "_____no_output_____" ], [ " The colour attribute for the red circle is the colour red, for the green circle object the colour attribute is green, and for the yellow circle the colour attribute is yellow. \n", "_____no_output_____" ], [ "<h3 id=\"method\">Methods</h3>", "_____no_output_____" ], [ "Methods give you a way to change or interact with the object; they are functions that interact with objects. For example, let’s say we would like to increase the radius by a specified amount of a circle. We can create a method called **add_radius(r)** that increases the radius by **r**. This is shown in figure 3, where after applying the method to the \"orange circle object\", the radius of the object increases accordingly. The “dot” notation means to apply the method to the object, which is essentially applying a function to the information in the object.", "_____no_output_____" ], [ "<img src=\"https://s3-api.us-geo.objectstorage.softlayer.net/cf-courses-data/CognitiveClass/PY0101EN/Chapter%203/Images/ClassesMethod.png\" width=\"500\" /> ", "_____no_output_____" ], [ "<i>Figure 3: Applying the method “add_radius” to the object orange circle object.</i>", "_____no_output_____" ], [ "<hr>", "_____no_output_____" ], [ "<h2 id=\"creating\">Creating a Class</h2>", "_____no_output_____" ], [ "Now we are going to create a class circle, but first, we are going to import a library to draw the objects: ", "_____no_output_____" ] ], [ [ "# Import the library\n\nimport matplotlib.pyplot as plt\n%matplotlib inline ", "_____no_output_____" ] ], [ [ " The first step in creating your own class is to use the <code>class</code> keyword, then the name of the class as shown in Figure 4. In this course the class parent will always be object: ", "_____no_output_____" ], [ "<img src=\"https://s3-api.us-geo.objectstorage.softlayer.net/cf-courses-data/CognitiveClass/PY0101EN/Chapter%203/Images/ClassesDefine.png\" width=\"400\" />", "_____no_output_____" ], [ "<i>Figure 4: Three instances of the class circle or three objects of type circle.</i>", "_____no_output_____" ], [ "The next step is a special method called a constructor <code>&#95;&#95;init&#95;&#95;</code>, which is used to initialize the object. The input are data attributes. The term <code>self</code> contains all the attributes in the set. For example the <code>self.color</code> gives the value of the attribute color and <code>self.radius</code> will give you the radius of the object. We also have the method <code>add_radius()</code> with the parameter <code>r</code>, the method adds the value of <code>r</code> to the attribute radius. To access the radius we use the syntax <code>self.radius</code>. The labeled syntax is summarized in Figure 5:", "_____no_output_____" ], [ "<img src=\"https://s3-api.us-geo.objectstorage.softlayer.net/cf-courses-data/CognitiveClass/PY0101EN/Chapter%203/Images/ClassesCircle.png\" width=\"600\" />", "_____no_output_____" ], [ "<i>Figure 5: Labeled syntax of the object circle.</i>", "_____no_output_____" ], [ "The actual object is shown below. We include the method <code>drawCircle</code> to display the image of a circle. We set the default radius to 3 and the default colour to blue:", "_____no_output_____" ] ], [ [ "# Create a class Circle\n\nclass Circle(object):\n \n # Constructor\n def __init__(self, radius=3, color='blue'):\n self.radius = radius\n self.color = color \n \n # Method\n def add_radius(self, r):\n self.radius = self.radius + r\n return(self.radius)\n \n # Method\n def drawCircle(self):\n plt.gca().add_patch(plt.Circle((0, 0), radius=self.radius, fc=self.color))\n plt.axis('scaled')\n plt.show() ", "_____no_output_____" ] ], [ [ "<hr>", "_____no_output_____" ], [ "<h2 id=\"circle\">Creating an instance of a class Circle</h2>", "_____no_output_____" ], [ "Let’s create the object <code>RedCircle</code> of type Circle to do the following:", "_____no_output_____" ] ], [ [ "# Create an object RedCircle\n\nRedCircle = Circle(10, 'red')", "_____no_output_____" ] ], [ [ "We can use the <code>dir</code> command to get a list of the object's methods. Many of them are default Python methods.", "_____no_output_____" ] ], [ [ "# Find out the methods can be used on the object RedCircle\n\ndir(RedCircle)", "_____no_output_____" ] ], [ [ "We can look at the data attributes of the object: ", "_____no_output_____" ] ], [ [ "# Print the object attribute radius\n\nRedCircle.radius", "_____no_output_____" ], [ "# Print the object attribute color\n\nRedCircle.color", "_____no_output_____" ] ], [ [ " We can change the object's data attributes: ", "_____no_output_____" ] ], [ [ "# Set the object attribute radius\n\nRedCircle.radius = 1\nRedCircle.radius", "_____no_output_____" ] ], [ [ " We can draw the object by using the method <code>drawCircle()</code>:", "_____no_output_____" ] ], [ [ "# Call the method drawCircle\n\nRedCircle.drawCircle()", "_____no_output_____" ] ], [ [ "We can increase the radius of the circle by applying the method <code>add_radius()</code>. Let increases the radius by 2 and then by 5: ", "_____no_output_____" ] ], [ [ "# Use method to change the object attribute radius\n\nprint('Radius of object:',RedCircle.radius)\nRedCircle.add_radius(2)\nprint('Radius of object of after applying the method add_radius(2):',RedCircle.radius)\nRedCircle.add_radius(5)\nprint('Radius of object of after applying the method add_radius(5):',RedCircle.radius)", "_____no_output_____" ] ], [ [ " Let’s create a blue circle. As the default colour is blue, all we have to do is specify what the radius is:", "_____no_output_____" ] ], [ [ "# Create a blue circle with a given radius\n\nBlueCircle = Circle(radius=100)", "_____no_output_____" ] ], [ [ " As before we can access the attributes of the instance of the class by using the dot notation:", "_____no_output_____" ] ], [ [ "# Print the object attribute radius\n\nBlueCircle.radius", "_____no_output_____" ], [ "# Print the object attribute color\n\nBlueCircle.color", "_____no_output_____" ] ], [ [ " We can draw the object by using the method <code>drawCircle()</code>:", "_____no_output_____" ] ], [ [ "# Call the method drawCircle\n\nBlueCircle.drawCircle()", "_____no_output_____" ] ], [ [ "Compare the x and y axis of the figure to the figure for <code>RedCircle</code>; they are different.", "_____no_output_____" ], [ "<hr>", "_____no_output_____" ], [ "<h2 id=\"rect\">The Rectangle Class</h2>", "_____no_output_____" ], [ "Let's create a class rectangle with the attributes of height, width and color. We will only add the method to draw the rectangle object:", "_____no_output_____" ] ], [ [ "# Create a new Rectangle class for creating a rectangle object\n\nclass Rectangle(object):\n \n # Constructor\n def __init__(self, width=2, height=3, color='r'):\n self.height = height \n self.width = width\n self.color = color\n \n # Method\n def drawRectangle(self):\n plt.gca().add_patch(plt.Rectangle((0, 0), self.width, self.height ,fc=self.color))\n plt.axis('scaled')\n plt.show()", "_____no_output_____" ] ], [ [ "Let’s create the object <code>SkinnyBlueRectangle</code> of type Rectangle. Its width will be 2 and height will be 3, and the color will be blue:", "_____no_output_____" ] ], [ [ "# Create a new object rectangle\n\nSkinnyBlueRectangle = Rectangle(2, 10, 'blue')", "_____no_output_____" ] ], [ [ " As before we can access the attributes of the instance of the class by using the dot notation:", "_____no_output_____" ] ], [ [ "# Print the object attribute height\n\nSkinnyBlueRectangle.height ", "_____no_output_____" ], [ "# Print the object attribute width\n\nSkinnyBlueRectangle.width", "_____no_output_____" ], [ "# Print the object attribute color\n\nSkinnyBlueRectangle.color", "_____no_output_____" ] ], [ [ " We can draw the object:", "_____no_output_____" ] ], [ [ "# Use the drawRectangle method to draw the shape\n\nSkinnyBlueRectangle.drawRectangle()", "_____no_output_____" ] ], [ [ "Let’s create the object <code>FatYellowRectangle</code> of type Rectangle :", "_____no_output_____" ] ], [ [ "# Create a new object rectangle\n\nFatYellowRectangle = Rectangle(20, 5, 'yellow')", "_____no_output_____" ] ], [ [ " We can access the attributes of the instance of the class by using the dot notation:", "_____no_output_____" ] ], [ [ "# Print the object attribute height\n\nFatYellowRectangle.height ", "_____no_output_____" ], [ "# Print the object attribute width\n\nFatYellowRectangle.width", "_____no_output_____" ], [ "# Print the object attribute color\n\nFatYellowRectangle.color", "_____no_output_____" ] ], [ [ " We can draw the object:", "_____no_output_____" ] ], [ [ "# Use the drawRectangle method to draw the shape\n\nFatYellowRectangle.drawRectangle()", "_____no_output_____" ] ], [ [ "<hr>\n<h2>The last exercise!</h2>\n<p>Congratulations, you have completed your first lesson and hands-on lab in Python. However, there is one more thing you need to do. The Data Science community encourages sharing work. The best way to share and showcase your work is to share it on GitHub. By sharing your notebook on GitHub you are not only building your reputation with fellow data scientists, but you can also show it off when applying for a job. Even though this was your first piece of work, it is never too early to start building good habits. So, please read and follow <a href=\"https://cognitiveclass.ai/blog/data-scientists-stand-out-by-sharing-your-notebooks/\" target=\"_blank\">this article</a> to learn how to share your work.\n<hr>", "_____no_output_____" ], [ "<div class=\"alert alert-block alert-info\" style=\"margin-top: 20px\">\n<h2>Get IBM Watson Studio free of charge!</h2>\n <p><a href=\"https://cocl.us/bottemNotebooksPython101Coursera\"><img src=\"https://s3-api.us-geo.objectstorage.softlayer.net/cf-courses-data/CognitiveClass/PY0101EN/Ad/BottomAd.png\" width=\"750\" align=\"center\"></a></p>\n</div>", "_____no_output_____" ], [ "<h3>About the Authors:</h3> \n<p><a href=\"https://www.linkedin.com/in/joseph-s-50398b136/\" target=\"_blank\">Joseph Santarcangelo</a> is a Data Scientist at IBM, and holds a PhD in Electrical Engineering. His research focused on using Machine Learning, Signal Processing, and Computer Vision to determine how videos impact human cognition. Joseph has been working for IBM since he completed his PhD.</p>", "_____no_output_____" ], [ "Other contributors: <a href=\"www.linkedin.com/in/jiahui-mavis-zhou-a4537814a\">Mavis Zhou</a>", "_____no_output_____" ], [ "<hr>", "_____no_output_____" ], [ "<p>Copyright &copy; 2018 IBM Developer Skills Network. This notebook and its source code are released under the terms of the <a href=\"https://cognitiveclass.ai/mit-license/\">MIT License</a>.</p>", "_____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" ]

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

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Despesas - Autorizações de Pagamento do Governo do Estado da Paraíba\n## De Janeiro/2021 a Junho/2021", "_____no_output_____" ] ], [ [ "# Instalação pacotes\n\n!pip install pandas\n!pip install PyMySQL\n!pip install SQLAlchemy", "_____no_output_____" ], [ "import pandas as pd", "_____no_output_____" ], [ "# Carregar CSVs em data frame do pandas\n\ndf1 = pd.read_csv('../data/pagamento_exercicio_2021_mes_1.csv', encoding='ISO-8859-1',sep=';')\ndf2 = pd.read_csv('../data/pagamento_exercicio_2021_mes_2.csv', encoding='ISO-8859-1',sep=';')\ndf3 = pd.read_csv('../data/pagamento_exercicio_2021_mes_3.csv', encoding='ISO-8859-1',sep=';')\ndf4 = pd.read_csv('../data/pagamento_exercicio_2021_mes_4.csv', encoding='ISO-8859-1',sep=';')\ndf5 = pd.read_csv('../data/pagamento_exercicio_2021_mes_5.csv', encoding='ISO-8859-1',sep=';')\ndf6 = pd.read_csv('../data/pagamento_exercicio_2021_mes_6.csv', encoding='ISO-8859-1',sep=';')", "_____no_output_____" ], [ "# Concatenar todos os dataframes\n\ndf = pd.concat([df1, df2, df3, df4, df5, df6])", "_____no_output_____" ] ], [ [ "## Realização de análises e transformações", "_____no_output_____" ] ], [ [ "# Exibir as colunas\n\ndf.columns", "_____no_output_____" ], [ "# Exibir quantidade de linhas e colunas\n\ndf.shape", "_____no_output_____" ], [ "# Exibir tipos das colunas\n\ndf.dtypes", "_____no_output_____" ], [ "# Converter coluna (DATA_PAGAMENTO) em datetime\n# Converter colunas (EXERCICIO, CODIGO_UNIDADE_GESTORA, NUMERO_EMPENHO, NUMERO_AUTORIZACAO_PAGAMENTO) em object\n\ndf[\"DATA_PAGAMENTO\"] = pd.to_datetime(df[\"DATA_PAGAMENTO\"])\ndf[\"EXERCICIO\"] = df[\"EXERCICIO\"].astype(\"object\")\ndf[\"CODIGO_UNIDADE_GESTORA\"] = df[\"CODIGO_UNIDADE_GESTORA\"].astype(\"object\")\ndf[\"NUMERO_EMPENHO\"] = df[\"CODIGO_UNIDADE_GESTORA\"].astype(\"object\")\ndf[\"NUMERO_AUTORIZACAO_PAGAMENTO\"] = df[\"NUMERO_AUTORIZACAO_PAGAMENTO\"].astype(\"object\")", "_____no_output_____" ], [ "# Exibir tipos das colunas\n\ndf.dtypes", "_____no_output_____" ], [ "# Consultar linhas com valores faltantes\n\ndf.isnull().sum()", "_____no_output_____" ], [ "# Exibir amostra\n\ndf.sample(10)", "_____no_output_____" ], [ "# Criar nova coluna que vai receber o mês de pagamento\n\ndf[\"MES_PAGAMENTO\"] = df[\"DATA_PAGAMENTO\"].dt.month", "_____no_output_____" ], [ "# Exibir amostra\n\ndf.sample(10)", "_____no_output_____" ], [ "# Conveter saída para coluna (VALOR_PAGAMENTO) com o tipo float\n\npd.options.display.float_format = 'R${:,.2f}'.format", "_____no_output_____" ], [ "# Retornar total pago agrupado por mês e por tipo de despesa\n\n# df.groupby([df[\"MES_PAGAMENTO\"], \"TIPO_DESPESA\"])[\"VALOR_PAGAMENTO\"].sum().reset_index()\n\n# Outra forma\ndf.groupby(['MES_PAGAMENTO', \"TIPO_DESPESA\"]).agg({\"VALOR_PAGAMENTO\":\"sum\"}).reset_index()", "_____no_output_____" ], [ "# Retornar maior valor pago a um credor agrupado por mês\n\n# df.groupby(df[\"MES_PAGAMENTO\"])[\"VALOR_PAGAMENTO\"].max()\n\ndf.groupby([\"MES_PAGAMENTO\"]).agg({\"VALOR_PAGAMENTO\":\"max\"}).reset_index()", "_____no_output_____" ], [ "# Salvar dataframe em um arquivo CSV\n\ndf.to_csv('../data/pagamento_exercicio_2021_jan_a_jun_governo_pb.csv', index=False)", "_____no_output_____" ], [ "# Salvar dataframe no banco de dados\n\nfrom sqlalchemy import create_engine\n\ncon = create_engine(\"mysql+pymysql://root:mysql@localhost:3307/db_governo_pb\",\n encoding=\"utf-8\")\ndf.to_sql('tb_pagamento_exercicio_2021', con, index = False, if_exists = 'replace', method = 'multi', chunksize=10000)", "_____no_output_____" ] ], [ [ "## Gráficos para análise exploratória e/ou tomada de decisão", "_____no_output_____" ] ], [ [ "import matplotlib.pyplot as plt\nplt.style.use(\"seaborn\")", "_____no_output_____" ], [ "# Gráfico com o total pago aos credores por mês (Janeiro a Junho)\n\ndf.groupby(df['MES_PAGAMENTO'])['VALOR_PAGAMENTO'].sum().plot.bar(title = 'Total Pago por Mês', color = 'blue')\nplt.xlabel('MÊS')\nplt.ylabel('RECEITA');", "_____no_output_____" ], [ "# Gráfico com o valor máximo pago a um credor por mês (Janeiro a Junho)\n\ndf.groupby([\"MES_PAGAMENTO\"]).agg({\"VALOR_PAGAMENTO\":\"max\"}).plot.bar(title = 'Maior valor pago a um credor po mês', color = 'green')\nplt.xlabel('MÊS')\nplt.ylabel('VALOR');", "_____no_output_____" ], [ "# Gráfico de linha exibindo a soma dos pagamentos a credores no decorrer dos meses\n\ndf.groupby([\"MES_PAGAMENTO\"]).agg({\"VALOR_PAGAMENTO\":\"sum\"}).plot(title = 'Total de pagamentos por mês aos credores')\nplt.xlabel('MÊS')\nplt.ylabel('TOTAL PAGO')\nplt.legend();", "_____no_output_____" ], [ "# Gráfico com o valor pago a credores agrupados por tipo de despesa\n\ndf.groupby([\"TIPO_DESPESA\"]).agg({\"VALOR_PAGAMENTO\":\"sum\"}).plot.bar(title = 'Soma dos valores pagos por tipo de despesa', color = 'gray')\nplt.xlabel('TIPO DE DESPESA')\nplt.ylabel('VALOR');", "_____no_output_____" ] ] ]

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

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

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "**Load the libraries:**", "_____no_output_____" ] ], [ [ "import numpy as np\nimport pandas as pd\n\nfrom sklearn.model_selection import train_test_split\n\nfrom keras.models import Sequential\nfrom keras.layers import Dense, Dropout\nfrom keras.callbacks import EarlyStopping, ModelCheckpoint\nfrom keras.optimizers import SGD, Adadelta, Adam, RMSprop, Adagrad, Nadam, Adamax\n\nSEED = 2017", "Using TensorFlow backend.\n" ] ], [ [ "**Import the dataset and extract the target variable:**", "_____no_output_____" ] ], [ [ "data = pd.read_csv('https://archive.ics.uci.edu/ml/machine-learning-databases/wine-quality/winequality-red.csv',\n sep = ';')\n\ny = data['quality']\nX = data.drop(['quality'], axis = 1)", "_____no_output_____" ] ], [ [ "**Split the dataset for training, validation and testing:**", "_____no_output_____" ] ], [ [ "X_train, X_test, y_train, y_test = train_test_split(X, y, \n test_size = 0.2,\n random_state = SEED)\n\nX_train, X_val, y_train, y_val = train_test_split(X_train, y_train,\n test_size = 0.2,\n random_state = SEED)", "_____no_output_____" ] ], [ [ "**Define a function that creates the model:**", "_____no_output_____" ] ], [ [ "def create_model(opt):\n model = Sequential()\n model.add(Dense(100, input_dim = X_train.shape[1],\n activation = 'relu'))\n model.add(Dense(50, activation = 'relu'))\n model.add(Dense(25, activation = 'relu'))\n model.add(Dense(10, activation = 'relu'))\n model.add(Dense(1, activation = 'linear'))\n return model", "_____no_output_____" ] ], [ [ "**Create a function that defines callbacks we will be using during training:**", "_____no_output_____" ] ], [ [ "def create_callbacks(opt):\n callbacks = [\n EarlyStopping(monitor = 'val_acc', patience = 200,\n verbose = 2),\n ModelCheckpoint('optimizers_best_' + opt + '.h5',\n monitor = 'val_acc',\n save_best_only = True,\n verbose = 0)\n ]\n return callbacks", "_____no_output_____" ] ], [ [ "**Create a dict of the optimizers we want to try:**", "_____no_output_____" ] ], [ [ "opts = dict({\n 'sgd': SGD(),\n 'sgd-0001': SGD(lr = 0.0001, decay = 0.00001),\n 'adam': Adam(),\n 'adadelta': Adadelta(),\n 'rmsprop': RMSprop(),\n 'rmsprop-0001': RMSprop(lr = 0.0001),\n 'nadam': Nadam(),\n 'adamax': Adamax()\n})", "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\nInstructions for updating:\nColocations handled automatically by placer.\n" ] ], [ [ "**Train our networks and store results:**", "_____no_output_____" ] ], [ [ "batch_size = 128\nn_epochs = 1000\n\nresults = []\n\n# Loop through the optimizers\nfor opt in opts:\n model = create_model(opt)\n callbacks = create_callbacks(opt)\n model.compile(loss = 'mse', \n optimizer = opts[opt],\n metrics = ['accuracy'])\n hist = model.fit(X_train.values, y_train, \n batch_size = batch_size,\n epochs = n_epochs,\n validation_data = (X_val.values, y_val),\n verbose = 0,\n callbacks = callbacks)\n \n best_epoch = np.argmax(hist.history['val_acc'])\n best_acc = hist.history['val_acc'][best_epoch]\n best_model = create_model(opt)\n \n # Load the model weights with the highest validation accuracy\n best_model.load_weights('optimizers_best_' + opt + '.h5')\n best_model.compile(loss = 'mse',\n optimizer = opts[opt],\n metrics = ['accuracy'])\n score = best_model.evaluate(X_test.values, y_test, verbose = 0)\n results.append([opt, best_epoch, best_acc, score[1]])", "Epoch 00201: early stopping\nEpoch 00414: early stopping\nEpoch 00625: early stopping\nEpoch 00373: early stopping\nEpoch 00413: early stopping\nEpoch 00230: early stopping\nEpoch 00269: early stopping\nEpoch 00424: early stopping\n" ] ], [ [ "**Compare the results:**", "_____no_output_____" ] ], [ [ "res = pd.DataFrame(results)\n\nres.columns = ['optimizer', 'epochs', 'val_accuracy', 'test_accuracy']\nres", "_____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" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Jupyter (IPython) Advanced Features\n---", "_____no_output_____" ], [ "Outline\n- Keyboard shortcuts\n- Magic\n- Accessing the underlying operating system\n- Using different languages inside single notebook\n- File magic\n- Using Jupyter more efficiently\n- Profiling\n- Output\n- Automation\n- Extensions\n- 'Big Data' Analysis\n \n\nSources: [IPython Tutorial](https://github.com/ipython/ipython-in-depth/blob/pycon-2019/1%20-%20Beyond%20Plain%20Python.ipynb), [Dataquest](https://www.dataquest.io/blog/advanced-jupyter-notebooks-tutorial/), and [Dataquest](https://www.dataquest.io/blog/jupyter-notebook-tips-tricks-shortcuts/), [Alex Rogozhnikov Blog](http://arogozhnikov.github.io/2016/09/10/jupyter-features.html) [Toward Data Science](https://towardsdatascience.com/how-to-effortlessly-optimize-jupyter-notebooks-e864162a06ee)", "_____no_output_____" ], [ "---", "_____no_output_____" ], [ "## Keyboard Shortcuts\n\n\nKeyboard Shortcuts\n\nAs in the classic Notebook, you can navigate the user interface through keyboard shortcuts. You can find and customize the current list of keyboard shortcuts by selecting the Advanced Settings Editor item in the Settings menu, then selecting Keyboard Shortcuts in the Settings tab.\n\n### Shortcut Keys for Jupyter lab\n\nWhile working with any tools, it helps if you know shortcut key to perform most frequent tasks. It increases your productivity and can be very comfortable while working. I have listed down some of the shortcuts which I use frequently while working on Jupyter Lab. Hopefully, it will be useful for others too. Also, you can check full list of shortcut by accessing the __commands tab__ in the Jupyter lab. You will find it below the Files on the left hand side.\n\n1. **ESC** takes users into command mode view while **ENTER** takes users into cell mode view.\n2. **A** inserts a cell above the currently selected cell. Before using this, make sure that you are in command mode (by pressing ESC).\n3. **B** inserts a cell below the currently selected cell. Before using this make sure that you are in command mode (by pressing ESC).\n4. **D + D** = Pressing D two times in a quick succession in command mode deletes the currently selected cell. \n5. Jupyter Lab gives you an option to change your cell into Code cell, Markdown cell or Raw Cell. You can use **M** to change current cell to a markdown cell, **Y** to change it to a code cell and  **R** to change it to a raw cell.\n6. ****CTRL + B**** = Jupyter lab has two columns design. One column is for launcher or code blocks and another column is for file view etc. To increase workspace while writing code, we can close it.  **CTRL + B** is the shortcut for toggling the file view column in the Jupyter lab.\n7. **SHIFT + M** = It merges multiple selected cells into one cell. \n8. **CTRL + SHIFT + –** = It splits the current cell into two cells from where your cursor is.\n9. **SHIFT+J** or **SHIFT + DOWN** = It selects the next cell in a downward direction.  It will help in making multiple selections of cells.\n10. **SHIFT + K** or **SHIFT + UP** = It selects the next cell in an upwards direction. It will help in making multiple selections of cells.\n11. **CTRL +** / = It helps you in either commenting or uncommenting any line in the Jupyter lab. For this to work, you don’t even need to select the whole line. It will comment or uncomment line where your cursor is. If you want to do it for more that one line then you will need to first select all the line and then use this shortcut.\n\nA PDF!!!\n- https://blog.ja-ke.tech/2019/01/20/jupyterlab-shortcuts.html\n- https://github.com/Jakeler/jupyter-shortcuts", "_____no_output_____" ], [ "## Magics\n\n---", "_____no_output_____" ], [ "Magics are turning simple python into *magical python*. Magics are the key to power of ipython.\n\nMagic functions are prefixed by % or %%, and typically take their arguments without parentheses, quotes or even commas for convenience. Line magics take a single % and cell magics are prefixed with two %%.", "_____no_output_____" ], [ "#### What is Magic??? Information about IPython's 'magic' % functions.", "_____no_output_____" ] ], [ [ "%magic", "_____no_output_____" ] ], [ [ "#### List available python magics", "_____no_output_____" ] ], [ [ "%lsmagic", "_____no_output_____" ] ], [ [ "#### %env\nYou can manage environment variables of your notebook without restarting the jupyter server process. Some libraries (like theano) use environment variables to control behavior, %env is the most convenient way.", "_____no_output_____" ] ], [ [ "# %env - without arguments lists environmental variables\n%env OMP_NUM_THREADS=4", "_____no_output_____" ] ], [ [ "# Accessing the underlying operating system\n\n---", "_____no_output_____" ], [ "## Executing shell commands\n\nYou can call any shell command. This in particular useful to manage your virtual environment.", "_____no_output_____" ] ], [ [ "!pip install numpy", "Requirement already satisfied: numpy in /Users/squiresrb/opt/anaconda3/lib/python3.8/site-packages (1.19.2)\n" ], [ "!pip list | grep Theano", "_____no_output_____" ] ], [ [ "## Adding packages can also be done using...", "_____no_output_____" ], [ "%conda install numpy", "_____no_output_____" ], [ "%pip install numpy", "_____no_output_____" ], [ "will attempt to install packages in the current environment.", "_____no_output_____" ] ], [ [ "!pwd", "/Users/squiresrb/Documents/GitHub/jupyter_training_2020/notebooks/2-Jupyter\n" ], [ "%pwd", "_____no_output_____" ], [ "pwd", "_____no_output_____" ], [ "files = !ls .\nprint(\"files in notebooks directory:\")\nprint(files)", "files in notebooks directory:\n['2-1-Jupyter-ecosystem.ipynb', '2-1-jupyter-ecosystem.slides.html', '2-10-jupyter-code-script-of-scripts.ipynb', '2-11-Advanced-jupyter.ipynb', '2-2-jupyter-get-in-and-out.ipynb', '2-3-jupyter-notebook-basics.ipynb', '2-4-jupyter-markdown.ipynb', '2-5-jupyter-code-python.ipynb', '2-6-jupyter-code-r.ipynb', '2-7-jupyter-command-line.ipynb', '2-8-jupyter-magics.ipynb', '2-9-jupyter-sharing', '2-Jupyter-help.ipynb', 'Advanced_jupyter.ipynb', 'big-data-analysis-jupyter.ipynb', 'foo.py', 'images', 'jupyter-advanced.ipynb', 'matplotlib-anatomy.ipynb', 'pythoncode.py']\n" ], [ "!echo $files", "[2-1-Jupyter-ecosystem.ipynb, 2-1-jupyter-ecosystem.slides.html, 2-10-jupyter-code-script-of-scripts.ipynb, 2-11-Advanced-jupyter.ipynb, 2-2-jupyter-get-in-and-out.ipynb, 2-3-jupyter-notebook-basics.ipynb, 2-4-jupyter-markdown.ipynb, 2-5-jupyter-code-python.ipynb, 2-6-jupyter-code-r.ipynb, 2-7-jupyter-command-line.ipynb, 2-8-jupyter-magics.ipynb, 2-9-jupyter-sharing, 2-Jupyter-help.ipynb, Advanced_jupyter.ipynb, big-data-analysis-jupyter.ipynb, foo.py, images, jupyter-advanced.ipynb, matplotlib-anatomy.ipynb, pythoncode.py]\n" ], [ "!echo {files[0].upper()}", "2-1-JUPYTER-ECOSYSTEM.IPYNB\n" ] ], [ [ "Note that all this is available even in multiline blocks:", "_____no_output_____" ] ], [ [ "import os\nfor i,f in enumerate(files):\n if f.endswith('ipynb'):\n !echo {\"%02d\" % i} - \"{os.path.splitext(f)[0]}\"\n else:\n print('--')", "00 - 2-1-Jupyter-ecosystem\n--\n02 - 2-10-jupyter-code-script-of-scripts\n03 - 2-11-Advanced-jupyter\n04 - 2-2-jupyter-get-in-and-out\n05 - 2-3-jupyter-notebook-basics\n06 - 2-4-jupyter-markdown\n07 - 2-5-jupyter-code-python\n08 - 2-6-jupyter-code-r\n09 - 2-7-jupyter-command-line\n10 - 2-8-jupyter-magics\n--\n12 - 2-Jupyter-help\n13 - Advanced_jupyter\n14 - big-data-analysis-jupyter\n--\n--\n17 - jupyter-advanced\n18 - matplotlib-anatomy\n--\n" ] ], [ [ "## I could take the same list with a bash command\n\nbecause magics and bash calls return python variables:", "_____no_output_____" ] ], [ [ "names = !ls ../images/ml_demonstrations/*.png\nnames[:5]", "_____no_output_____" ] ], [ [ "## Suppress output of last line\n\nsometimes output isn't needed, so we can either use `pass` instruction on new line or semicolon at the end ", "_____no_output_____" ], [ "%conda install matplotlib", "_____no_output_____" ] ], [ [ "%matplotlib inline\nfrom matplotlib import pyplot as plt\nimport numpy", "_____no_output_____" ], [ "# if you don't put semicolon at the end, you'll have output of function printed\n\nplt.hist(numpy.linspace(0, 1, 1000)**1.5);", "_____no_output_____" ] ], [ [ "# Using different languages inside single notebook\n\n---\n\nIf you're missing those much, using other computational kernels:\n\n- %%python2\n- %%python3\n- %%ruby\n- %%perl\n- %%bash\n- %%R\n\nis possible, but obviously you'll need to setup the corresponding kernel first.", "_____no_output_____" ] ], [ [ "# %%ruby\n# puts 'Hi, this is ruby.'", "_____no_output_____" ], [ "%%bash\necho 'Hi, this is bash.'", "Hi, this is bash.\n" ] ], [ [ "## Running R code in Jupyter notebook", "_____no_output_____" ], [ "#### Installing R kernel\n\nEasy Option: Installing the R Kernel Using Anaconda\nIf you used Anaconda to set up your environment, getting R working is extremely easy. Just run the below in your terminal:", "_____no_output_____" ] ], [ [ "# %conda install -c r r-essentials", "_____no_output_____" ] ], [ [ "#### Running R and Python in the same notebook.\n\nThe best solution to this is to install rpy2 (requires a working version of R as well), which can be easily done with pip:", "_____no_output_____" ] ], [ [ "%pip install rpy2", "Collecting rpy2\n Downloading rpy2-3.3.6.tar.gz (179 kB)\n\u001b[K |████████████████████████████████| 179 kB 465 kB/s eta 0:00:01\n\u001b[31m ERROR: Command errored out with exit status 1:\n command: /Users/squiresrb/opt/anaconda3/bin/python -c 'import sys, setuptools, tokenize; sys.argv[0] = '\"'\"'/private/var/folders/d7/8vn6rd1d6f37gtgy_13h3b95mx3fmv/T/pip-install-79ynf4cx/rpy2/setup.py'\"'\"'; __file__='\"'\"'/private/var/folders/d7/8vn6rd1d6f37gtgy_13h3b95mx3fmv/T/pip-install-79ynf4cx/rpy2/setup.py'\"'\"';f=getattr(tokenize, '\"'\"'open'\"'\"', open)(__file__);code=f.read().replace('\"'\"'\\r\\n'\"'\"', '\"'\"'\\n'\"'\"');f.close();exec(compile(code, __file__, '\"'\"'exec'\"'\"'))' egg_info --egg-base /private/var/folders/d7/8vn6rd1d6f37gtgy_13h3b95mx3fmv/T/pip-pip-egg-info-8nhbgt8z\n cwd: /private/var/folders/d7/8vn6rd1d6f37gtgy_13h3b95mx3fmv/T/pip-install-79ynf4cx/rpy2/\n Complete output (2 lines):\n cffi mode: CFFI_MODE.ANY\n Error: rpy2 in API mode cannot be built without R in the PATH or R_HOME defined. Correct this or force ABI mode-only by defining the environment variable RPY2_CFFI_MODE=ABI\n ----------------------------------------\u001b[0m\n\u001b[31mERROR: Command errored out with exit status 1: python setup.py egg_info Check the logs for full command output.\u001b[0m\n\u001b[?25hNote: you may need to restart the kernel to use updated packages.\n" ] ], [ [ "You can then use the two languages together, and even pass variables inbetween:", "_____no_output_____" ] ], [ [ "%load_ext rpy2.ipython", "_____no_output_____" ], [ "%R require(ggplot2)", "_____no_output_____" ], [ "import pandas as pd\ndf = pd.DataFrame({\n 'Letter': ['a', 'a', 'a', 'b', 'b', 'b', 'c', 'c', 'c'],\n 'X': [4, 3, 5, 2, 1, 7, 7, 5, 9],\n 'Y': [0, 4, 3, 6, 7, 10, 11, 9, 13],\n 'Z': [1, 2, 3, 1, 2, 3, 1, 2, 3]\n })", "_____no_output_____" ], [ "%%R -i df\nggplot(data = df) + geom_point(aes(x = X, y= Y, color = Letter, size = Z))", "_____no_output_____" ] ], [ [ "## Writing functions in cython (or fortran)\n\nSometimes the speed of numpy is not enough and I need to write some fast code. \nIn principle, you can compile function in the dynamic library and write python wrappers...\n\nBut it is much better when this boring part is done for you, right?\n\nYou can write functions in cython or fortran and use those directly from python code.\n\nFirst you'll need to install:\n```\n%pip install cython \n```", "_____no_output_____" ] ], [ [ "%pip install cython", "_____no_output_____" ], [ "%load_ext Cython", "_____no_output_____" ], [ "%%cython\ndef myltiply_by_2(float x):\n return 2.0 * x", "_____no_output_____" ], [ "myltiply_by_2(23.)", "_____no_output_____" ] ], [ [ "I also should mention that there are different jitter systems which can speed up your python code.\nMore examples in [my notebook](http://arogozhnikov.github.io/2015/09/08/SpeedBenchmarks.html). \n", "_____no_output_____" ], [ "For more information see the IPython help at: [Cython](https://github.com/ipython/ipython-in-depth/blob/pycon-2019/6%20-%20Cross-Language-Integration.ipynb)", "_____no_output_____" ], [ "# File magic", "_____no_output_____" ], [ "%%writefile Export the contents of a cell", "_____no_output_____" ] ], [ [ "%%writefile?", "_____no_output_____" ] ], [ [ "`%pycat` ill output in the pop-up window:\n```\nShow a syntax-highlighted file through a pager.\n\nThis magic is similar to the cat utility, but it will assume the file\nto be Python source and will show it with syntax highlighting.\n\nThis magic command can either take a local filename, an url,\nan history range (see %history) or a macro as argument ::\n\n%pycat myscript.py\n%pycat 7-27\n%pycat myMacro\n%pycat http://www.example.com/myscript.py\n```", "_____no_output_____" ], [ "## %load \nloading code directly into cell. You can pick local file or file on the web.\n\nAfter uncommenting the code below and executing, it will replace the content of cell with contents of file.\n", "_____no_output_____" ] ], [ [ "# %load https://matplotlib.org/_downloads/f7171577b84787f4b4d987b663486a94/anatomy.py", "_____no_output_____" ] ], [ [ "## %run to execute python code\n\n%run can execute python code from .py files &mdash; this is a well-documented behavior. \n\nBut it also can execute other jupyter notebooks! Sometimes it is quite useful.\n\nNB. %run is not the same as importing python module.", "_____no_output_____" ] ], [ [ "# this will execute all the code cells from different notebooks\n%run ./matplotlib-anatomy.ipynb", "_____no_output_____" ] ], [ [ "# Using Jupyter more efficiently\n\n---", "_____no_output_____" ], [ "## Store Magic - %store: lazy passing data between notebooks", "_____no_output_____" ], [ "%store lets you store your macro and use it across all of your Jupyter Notebooks.", "_____no_output_____" ] ], [ [ "data = 'this is the string I want to pass to different notebook'\n%store data\ndel data # deleted variable", "_____no_output_____" ], [ "# in second notebook I will use:\n%store -r data\nprint(data)", "_____no_output_____" ] ], [ [ "## %who: analyze variables of global scope", "_____no_output_____" ] ], [ [ "%whos", "_____no_output_____" ], [ "# pring names of string variables\n%who str", "_____no_output_____" ] ], [ [ "## Multiple cursors\n\nSince recently jupyter supports multiple cursors (in a single cell), just like sublime ot intelliJ! __Alt + mouse selection__ for multiline selection and __Ctrl + mouse clicks__ for multicursors.\n\n<img src='./images/jupyter/multi-cursor.gif' />\n\nGif taken from http://swanintelligence.com/multi-cursor-in-jupyter.html", "_____no_output_____" ], [ "## Timing ", "_____no_output_____" ], [ "When you need to measure time spent or find the bottleneck in the code, ipython comes to the rescue.", "_____no_output_____" ], [ "%%time\nimport time\ntime.sleep(2) # sleep for two seconds", "_____no_output_____" ] ], [ [ "# measure small code snippets with timeit !\nimport numpy\n%timeit numpy.random.normal(size=100)", "_____no_output_____" ], [ "%%writefile pythoncode.py\n\nimport numpy\ndef append_if_not_exists(arr, x):\n if x not in arr:\n arr.append(x)\n \ndef some_useless_slow_function():\n arr = list()\n for i in range(10000):\n x = numpy.random.randint(0, 10000)\n append_if_not_exists(arr, x)", "_____no_output_____" ], [ "# shows highlighted source of the newly-created file\n%pycat pythoncode.py", "_____no_output_____" ], [ "from pythoncode import some_useless_slow_function, append_if_not_exists", "_____no_output_____" ] ], [ [ "## Hiding code or output", "_____no_output_____" ], [ "- Click on the blue vertical bar or line to the left to collapse code or output", "_____no_output_____" ], [ "## Commenting and uncommenting a block of code\n\nYou might want to add new lines of code and comment out the old lines while you’re working. This is great if you’re improving the performance of your code or trying to debug it.\n- First, select all the lines you want to comment out.\n- Next hit cmd + / to comment out the highlighted code!", "_____no_output_____" ], [ "## Pretty Print all cell outputs\n\nNormally only the last output in the cell will be printed. For everything else, you have to manually add print(), which is fine but not super convenient. You can change that by adding this at the top of the notebook:", "_____no_output_____" ] ], [ [ "from IPython.core.interactiveshell import InteractiveShell\nInteractiveShell.ast_node_interactivity = \"all\"", "_____no_output_____" ] ], [ [ "# Profiling: %prun, %lprun, %mprun\n---", "_____no_output_____" ], [ "See a much longer explination of profiling and timeing in Jake Vanderplas' Python Data Science Handbook: \nhttps://jakevdp.github.io/PythonDataScienceHandbook/01.07-timing-and-profiling.html", "_____no_output_____" ] ], [ [ "# shows how much time program spent in each function\n%prun some_useless_slow_function()", "_____no_output_____" ] ], [ [ "Example of output:\n```\n26338 function calls in 0.713 seconds\n\n Ordered by: internal time\n\n ncalls tottime percall cumtime percall filename:lineno(function)\n 10000 0.684 0.000 0.685 0.000 pythoncode.py:3(append_if_not_exists)\n 10000 0.014 0.000 0.014 0.000 {method 'randint' of 'mtrand.RandomState' objects}\n 1 0.011 0.011 0.713 0.713 pythoncode.py:7(some_useless_slow_function)\n 1 0.003 0.003 0.003 0.003 {range}\n 6334 0.001 0.000 0.001 0.000 {method 'append' of 'list' objects}\n 1 0.000 0.000 0.713 0.713 <string>:1(<module>)\n 1 0.000 0.000 0.000 0.000 {method 'disable' of '_lsprof.Profiler' objects}\n```", "_____no_output_____" ] ], [ [ "# %load_ext memory_profiler ???", "_____no_output_____" ], [ "To profile memory, you can install and run pmrun\n\n# %pip install memory_profiler\n# %pip install line_profiler", "_____no_output_____" ], [ "# tracking memory consumption (show in the pop-up)\n# %mprun -f append_if_not_exists some_useless_slow_function()", "_____no_output_____" ] ], [ [ "Example of output:\n```\nLine # Mem usage Increment Line Contents\n================================================\n 3 20.6 MiB 0.0 MiB def append_if_not_exists(arr, x):\n 4 20.6 MiB 0.0 MiB if x not in arr:\n 5 20.6 MiB 0.0 MiB arr.append(x)\n```", "_____no_output_____" ], [ "**%lprun** is line profiling, but it seems to be broken for latest IPython release, so we'll manage without magic this time:", "_____no_output_____" ], [ "```python\nimport line_profiler\nlp = line_profiler.LineProfiler()\nlp.add_function(some_useless_slow_function)\nlp.runctx('some_useless_slow_function()', locals=locals(), globals=globals())\nlp.print_stats()\n```", "_____no_output_____" ], [ "## Debugging with %debug\n\nJupyter has own interface for [ipdb](https://docs.python.org/2/library/pdb.html). Makes it possible to go inside the function and investigate what happens there.\n\nThis is not pycharm and requires much time to adapt, but when debugging on the server this can be the only option (or use pdb from terminal).", "_____no_output_____" ] ], [ [ "#%%debug filename:line_number_for_breakpoint\n# Here some code that fails. This will activate interactive context for debugging", "_____no_output_____" ] ], [ [ "A bit easier option is `%pdb`, which activates debugger when exception is raised:", "_____no_output_____" ] ], [ [ "# %pdb\n\n# def pick_and_take():\n# picked = numpy.random.randint(0, 1000)\n# raise NotImplementedError()\n \n# pick_and_take()", "_____no_output_____" ] ], [ [ "# Output\n---", "_____no_output_____" ], [ "## [RISE](https://github.com/damianavila/RISE): presentations with notebook\n\nExtension by Damian Avila makes it possible to show notebooks as demonstrations. Example of such presentation: http://bollwyvl.github.io/live_reveal/#/7\n\nIt is very useful when you teach others e.g. to use some library.\n", "_____no_output_____" ], [ "## Jupyter output system\n\nNotebooks are displayed as HTML and the cell output can be HTML, so you can return virtually anything: video/audio/images. \n\nIn this example I scan the folder with images in my repository and show first five of them:", "_____no_output_____" ] ], [ [ "import os\nfrom IPython.display import display, Image\nnames = [f for f in os.listdir('../images/') if f.endswith('.png')]\nfor name in names[:5]:\n display(Image('../images/' + name, width=300))", "_____no_output_____" ] ], [ [ "## Write your posts in notebooks\n\nLike this one. Use `nbconvert` to export them to html.", "_____no_output_____" ], [ "# [Jupyter-contrib extensions](https://github.com/ipython-contrib/jupyter_contrib_nbextensions)\n\nare installed with \n```\n!pip install https://github.com/ipython-contrib/jupyter_contrib_nbextensions/tarball/master\n!pip install jupyter_nbextensions_configurator\n!jupyter contrib nbextension install --user\n!jupyter nbextensions_configurator enable --user\n```\n\n<img src='./images/jupyter/nbextensions.png' />\n\nthis is a family of different extensions, including e.g. **jupyter spell-checker and code-formatter**, \nthat are missing in jupyter by default. ", "_____no_output_____" ], [ "## Reconnect to kernel\n\nLong before, when you started some long-taking process and at some point your connection to ipython server dropped, \nyou completely lost the ability to track the computations process (unless you wrote this information to file). So either you interrupt the kernel and potentially lose some progress, or you wait till it completes without any idea of what is happening.\n\n`Reconnect to kernel` option now makes it possible to connect again to running kernel without interrupting computations and get the newcoming output shown (but some part of output is already lost).", "_____no_output_____" ], [ "# Big data analysis\n\nA number of solutions are available for querying/processing large data samples: \n- [ipyparallel (formerly ipython cluster)](https://github.com/ipython/ipyparallel) is a good option for simple map-reduce operations in python. We use it in [rep](github.com/yandex/rep) to train many machine learning models in parallel\n- [pyspark](http://www.cloudera.com/documentation/enterprise/5-5-x/topics/spark_ipython.html)\n- spark-sql magic [%%sql](https://github.com/jupyter-incubator/sparkmagic)", "_____no_output_____" ], [ "Additional Resources:\n\n* IPython [built-in magics](https://ipython.org/ipython-doc/3/interactive/magics.html)\n* Nice [interactive presentation about jupyter](http://quasiben.github.io/dfwmeetup_2014/#/) by Ben Zaitlen\n* Advanced notebooks [part 1: magics](https://blog.dominodatalab.com/lesser-known-ways-of-using-notebooks/) and [part 2: widgets](https://blog.dominodatalab.com/interactive-dashboards-in-jupyter/)\n* [Profiling in python with jupyter](http://pynash.org/2013/03/06/timing-and-profiling/)\n* [4 ways to extend notebooks](http://mindtrove.info/4-ways-to-extend-jupyter-notebook/)\n* [IPython notebook tricks](https://www.quora.com/What-are-your-favorite-tricks-for-IPython-Notebook)\n* [Jupyter vs Zeppelin for big data](https://www.linkedin.com/pulse/comprehensive-comparison-jupyter-vs-zeppelin-hoc-q-phan-mba-)\n* [Making publication ready Python notebooks](http://blog.juliusschulz.de/blog/ultimate-ipython-notebook).\n* https://yoursdata.net/installing-and-configuring-jupyter-lab-on-windows/", "_____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" ]

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

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "<a href=\"https://colab.research.google.com/github/michaelll22/CPEN-21A-ECE-2-2/blob/main/Operations_and_Expressions_in_Python.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>", "_____no_output_____" ], [ "##Boolean Operators", "_____no_output_____" ] ], [ [ "a = 10\nb = 9\nc = 8\n\nprint (10 > 9)\nprint (10 == 9)\nprint (10 < 9)\n\nprint (a)\n\nprint (a > b)\nc = print (a > b)\n\nc", "True\nFalse\nFalse\n10\nTrue\nTrue\n" ], [ "##true\nprint(bool(\"Hello\"))\nprint(bool(15))\nprint(bool(True))\nprint(bool(1))\n\n##false\nprint(bool(False))\nprint(bool(0))\nprint(bool(None))\nprint(bool([]))", "True\nTrue\nTrue\nFalse\nTrue\nFalse\nFalse\nFalse\n" ], [ "def myFunction(): \n return True\n\nprint(myFunction())", "True\n" ], [ "def myFunction():\n return True\n\nif myFunction():\n print(\"Yes/True\")\nelse:\n print(\"No/Flase\")\n", "Yes\n" ], [ "print(10>9)\n\na = 6 #0000 0110\nb = 7 #0000 0111\n\nprint(a == b)\nprint(a != b)", "True\nFalse\nTrue\n" ] ], [ [ "##Python Operators", "_____no_output_____" ] ], [ [ "print(10 + 5)\nprint(10 - 5)\nprint(10 * 5)\nprint(10 / 5)\nprint(10 % 5)\nprint(10 // 3)\nprint(10 ** 2)", "15\n5\n50\n2.0\n0\n3\n100\n" ] ], [ [ "##Bitwise Operators", "_____no_output_____" ] ], [ [ "a = 60 #0011 1100\nb = 13 \n\nprint (a^b)\nprint (~a)\nprint (a<<2)\nprint (a>>2) #0000 1111", "49\n-61\n240\n15\n" ] ], [ [ "##Assignment Operator", "_____no_output_____" ] ], [ [ "x = 2\nx += 3 #Same As x = x+3\nprint(x)\nx", "5\n" ] ], [ [ "##Logical Operators", "_____no_output_____" ] ], [ [ "a = 5\nb = 6\n\nprint(a>b and a==a)\nprint(a<b or b==a)", "False\nTrue\n" ] ], [ [ "##Identity Operator", "_____no_output_____" ] ], [ [ "print(a is b)\nprint(a is not b)", "False\nTrue\n" ] ] ]

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

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

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import matplotlib.pyplot as plt\nimport pandas as pd\nimport numpy as np\nimport sys\nfrom time import time\n\n%pylab inline\npylab.rcParams['figure.figsize'] = (20.0, 10.0)\n\n%load_ext autoreload\n%autoreload 2\n\nsys.path.append('..')\n\nimport isolation\nimport sample_players\nimport run_match\nimport my_baseline_player as custom\nimport book as b", "Populating the interactive namespace from numpy and matplotlib\nThe autoreload extension is already loaded. To reload it, use:\n %reload_ext autoreload\n" ] ], [ [ "## I estimate 10s per game. 100 starting positions, 100 secondary starting positions, then 10000 openings. 4 threads, and symmetries that produce x4 data. If I want 12 points per opening, then that would be:", "_____no_output_____" ] ], [ [ "estimated_seconds = 10000 * 12 * 10/ (4 * 4)\nestimated_hours = estimated_seconds / 3600\nprint(estimated_hours)", "20.833333333333332\n" ] ], [ [ "### The plan is as follows:\n - Create a book (or load previously saved)\n - for each starting action for player 1 (100) and each starting action for player 2 (100) run 3 experiments (DETERMINISTIC BOOK FILLING).\n - Run epsilon-greedy algorithm to make a STOCHASTIC BOOK FILLING (using the opening book up to its depth [1-epsilon of the time]). Reduce epsilon exponentially to zero.", "_____no_output_____" ] ], [ [ "book = b.load_latest_book(depth=4)", "_____no_output_____" ], [ "type(book)", "_____no_output_____" ], [ "sum(abs(value) for value in book.values())", "_____no_output_____" ], [ "#book # book -> {(state, action): counts}", "_____no_output_____" ], [ "agent_names = ('CustomPlayer1', 'CustomPlayer2')\nagent1 = isolation.Agent(custom.CustomPlayer, agent_names[0])\nagent2 = isolation.Agent(custom.CustomPlayer, agent_names[1])\nagents = (agent1, agent2)\n\nstate = isolation.isolation.Isolation()\ntime_limit = 150\nmatch_id = 0\n\ntic = time.time()\nwinner, game_history, match_id = isolation.play((agents,\n state,\n time_limit,\n match_id))\ntoc = time.time()\nprint('Elapsed time: {}'.format((toc-tic)))", "Elapsed time: 11.986933946609497\n" ], [ "root = isolation.isolation.Isolation()\nopening_states = list(b.get_full_states(root, depth=2))\nprint(type(opening_states))\nprint(len(opening_states))", "<class 'list'>\n9801\n" ], [ "len([s for s in opening_states if s.ply_count==1])", "_____no_output_____" ], [ "[s for s in opening_states if s.ply_count==0]", "_____no_output_____" ], [ "99*99", "_____no_output_____" ], [ "opening_states[0]", "_____no_output_____" ] ], [ [ "### Let's generate the corresponding matches", "_____no_output_____" ] ], [ [ "# Constant parameteres\ntime_limit = 150\ndepth = 4\nfull_search_depth = 2\nmatches_per_opening = 3\n\n# Create the agents that will play\nagent_names = ('CustomPlayer1', 'CustomPlayer2')\nagent1 = isolation.Agent(custom.CustomPlayer, agent_names[0])\nagent2 = isolation.Agent(custom.CustomPlayer, agent_names[1])\nagents = (agent1, agent2)\n\n# Get the initial states\nroot = isolation.isolation.Isolation()\nopening_states = list(b.get_full_states(root, depth=full_search_depth))\n\n# Generate the matches\nmatches = [(agents, state, time_limit, match_id) \n for match_id, state in enumerate(opening_states)]\nmatches = matches * 3\nprint('Generated {} matches.'.format(len(matches)))\n\n# Create or load the book\nbook = b.load_latest_book(depth=depth)", "Generated 29403 matches.\n" ], [ "matches[0]", "_____no_output_____" ], [ "def active_player(state):\n return state.ply_count % 2", "_____no_output_____" ], [ "active_player(matches[0][1])", "_____no_output_____" ], [ "batch_size = 10\nx = list(range(10,45))\nbatches = [x[i*batch_size:(i+1)*batch_size] \n for i in range(len(x) // batch_size + (len(x) % batch_size != 0))]\nbatches", "_____no_output_____" ], [ "l = [1,2,3,445]", "_____no_output_____" ], [ "isinstance(l[3], int)", "_____no_output_____" ], [ "l.insert(0,45)\nl", "_____no_output_____" ], [ "from multiprocessing.pool import ThreadPool as Pool\nnum_processes = 1\nbatch_size = 10\n\n# Small test for debugging\nmatches = matches[:10]\n\nresults = []\npool = Pool(num_processes)\ntic = time.time()\nfor result in pool.imap_unordered(isolation.play, matches):\n results.append(result)\n winner, game_history, match_id = result\n print('Results for match {}: {} wins.'.format(match_id, winner.name))\n _, state, _, _ = matches[match_id]\n if state.locs[1] is not None:\n game_history.insert(0,state.locs[1])\n if state.locs[0] is not None:\n game_history.insert(0,state.locs[0])\n root = isolation.isolation.Isolation()\n print(game_history)\n b.process_game_history(root,\n game_history, \n book,\n agent_names.index(winner.name),\n active_player=state.ply_count % 2,\n depth=depth)\ntoc = time.time()\nprint('Elapsed time {} seconds.'.format((toc-tic)))", "Results for match 0: CustomPlayer1 wins.\n[84, 56, <Action.ENE: 11>, <Action.NNE: 25>, <Action.SSW: -25>, <Action.ENE: 11>, <Action.ESE: -15>, <Action.SSE: -27>, <Action.WSW: -11>, <Action.WSW: -11>, <Action.NNE: 25>, <Action.NNE: 25>, <Action.SSE: -27>, <Action.SSE: -27>, <Action.WSW: -11>, <Action.WSW: -11>, <Action.NNW: 27>, <Action.WSW: -11>, <Action.WSW: -11>, <Action.SSW: -25>, <Action.NNE: 25>, <Action.ENE: 11>, <Action.NNE: 25>, <Action.NNW: 27>, <Action.ESE: -15>, <Action.SSW: -25>, <Action.SSW: -25>, <Action.NNW: 27>, <Action.SSW: -25>, <Action.SSW: -25>, <Action.WSW: -11>, <Action.WNW: 15>, <Action.NNE: 25>, <Action.NNE: 25>, <Action.NNE: 25>, <Action.SSE: -27>, <Action.NNE: 25>, <Action.SSE: -27>, <Action.ENE: 11>, <Action.ENE: 11>, <Action.SSE: -27>, <Action.ESE: -15>, <Action.NNE: 25>, <Action.NNW: 27>, <Action.WSW: -11>, <Action.ENE: 11>, <Action.WNW: 15>, <Action.SSW: -25>, <Action.WSW: -11>, <Action.WSW: -11>, <Action.SSW: -25>, <Action.WNW: 15>, <Action.SSW: -25>, <Action.WSW: -11>, <Action.ENE: 11>, <Action.WNW: 15>, <Action.SSW: -25>]\nState: Isolation(board=41523161203939122082683632224299007, ply_count=0, locs=(None, None)) \n Action: 84\n\n\nGot an int action: loc_sym = 82\nGot an int action: loc_sym = 32\nGot an int action: loc_sym = 30\nState: Isolation(board=41523161184596308968849565429000191, ply_count=1, locs=(84, None)) \n Action: 56\n\n\nGot an int action: loc_sym = 58\nGot an int action: loc_sym = 56\nGot an int action: loc_sym = 58\nState: Isolation(board=41523161184596308896791971391072255, ply_count=2, locs=(84, 56)) \n Action: 11\n\n\nState: Isolation(board=41523121570515051764623174619097087, ply_count=3, locs=(95, 56)) \n Action: 25\n\n\nResults for match 1: CustomPlayer2 wins.\n[18, 79, <Action.NNE: 25>, <Action.SSE: -27>, <Action.NNE: 25>, <Action.WSW: -11>, <Action.NNE: 25>, <Action.NNE: 25>, <Action.ESE: -15>, <Action.WSW: -11>, <Action.SSW: -25>, <Action.WSW: -11>, <Action.WSW: -11>, <Action.SSW: -25>, <Action.WNW: 15>, <Action.NNW: 27>, <Action.NNE: 25>, <Action.ESE: -15>, <Action.WSW: -11>, <Action.NNE: 25>, <Action.NNE: 25>, <Action.WSW: -11>, <Action.SSE: -27>, <Action.NNE: 25>, <Action.WSW: -11>, <Action.WSW: -11>, <Action.NNE: 25>, <Action.WSW: -11>, <Action.WSW: -11>, <Action.NNE: 25>, <Action.WNW: 15>, <Action.ENE: 11>, <Action.NNE: 25>, <Action.NNE: 25>, <Action.ESE: -15>, <Action.ESE: -15>, <Action.ENE: 11>, <Action.ENE: 11>, <Action.SSE: -27>, <Action.SSW: -25>, <Action.SSE: -27>, <Action.WNW: 15>, <Action.ENE: 11>, <Action.WNW: 15>, <Action.SSW: -25>, <Action.SSW: -25>, <Action.SSE: -27>, <Action.SSW: -25>, <Action.WSW: -11>, <Action.SSW: -25>, <Action.NNE: 25>, <Action.ESE: -15>, <Action.WSW: -11>, <Action.NNW: 27>, <Action.ESE: -15>, <Action.SSW: -25>, <Action.NNE: 25>, <Action.ENE: 11>, <Action.WSW: -11>, <Action.SSE: -27>, <Action.WSW: -11>, <Action.ENE: 11>, <Action.NNE: 25>, <Action.ENE: 11>, <Action.ESE: -15>, <Action.SSW: -25>, <Action.NNE: 25>]\nState: Isolation(board=41523161203939122082683632224299007, ply_count=0, locs=(None, None)) \n Action: 18\n\n\nGot an int action: loc_sym = 18\nGot an int action: loc_sym = 96\nGot an int action: loc_sym = 96\nState: Isolation(board=41523161203939122082683632224036863, ply_count=1, locs=(18, None)) \n Action: 79\n\n\nGot an int action: loc_sym = 87\nGot an int action: loc_sym = 27\nGot an int action: loc_sym = 35\nState: Isolation(board=41523161203334659172876317636683775, ply_count=2, locs=(18, 79)) \n Action: 25\n\n\nState: Isolation(board=41523161203334659172867521543661567, ply_count=3, locs=(43, 79)) \n Action: -27\n\n\n\nResults for match 2: CustomPlayer1 wins.\n[81, 29, <Action.NNE: 25>, <Action.NNE: 25>, <Action.ESE: -15>, <Action.NNE: 25>, <Action.SSW: -25>, <Action.SSE: -27>, <Action.SSW: -25>, <Action.SSW: -25>, <Action.SSW: -25>, <Action.WNW: 15>, <Action.WNW: 15>, <Action.NNE: 25>, <Action.NNE: 25>, <Action.ENE: 11>, <Action.WSW: -11>, <Action.SSW: -25>, <Action.NNE: 25>, <Action.SSW: -25>, <Action.WSW: -11>, <Action.WNW: 15>, <Action.NNE: 25>, <Action.WNW: 15>, <Action.NNE: 25>, <Action.NNE: 25>, <Action.SSE: -27>, <Action.WSW: -11>, <Action.WSW: -11>, <Action.ESE: -15>, <Action.NNE: 25>, <Action.SSE: -27>, <Action.WSW: -11>, <Action.NNE: 25>, <Action.SSW: -25>, <Action.ESE: -15>, <Action.SSE: -27>, <Action.NNE: 25>, <Action.ENE: 11>, <Action.WNW: 15>, <Action.SSW: -25>, <Action.WNW: 15>, <Action.NNW: 27>, <Action.WNW: 15>, <Action.SSW: -25>, <Action.WSW: -11>, <Action.ENE: 11>, <Action.SSW: -25>, <Action.WNW: 15>, <Action.NNW: 27>, <Action.SSE: -27>, <Action.ENE: 11>, <Action.WSW: -11>, <Action.ESE: -15>, <Action.NNE: 25>, <Action.ENE: 11>, <Action.SSE: -27>, <Action.ESE: -15>, <Action.ENE: 11>, <Action.ENE: 11>]\nState: Isolation(board=41523161203939122082683632224299007, ply_count=0, locs=(None, None)) \n Action: 81\n\n\nGot an int action: loc_sym = 85\nGot an int action: loc_sym = 29\nGot an int action: loc_sym = 33\nState: Isolation(board=41523161201521270443454373874886655, ply_count=1, locs=(81, None)) \n Action: 29\n\n\nGot an int action: loc_sym = 33\nGot an int action: loc_sym = 81\nGot an int action: loc_sym = 85\nState: Isolation(board=41523161201521270443454373338015743, ply_count=2, locs=(81, 29)) \n Action: 25\n\n\nState: Isolation(board=41442031563106663761758584332871679, ply_count=3, locs=(106, 29)) \n Action: 25\n\n\nResults for match 3: CustomPlayer2 wins.\n[108, 2, <Action.ESE: -15>, <Action.NNE: 25>, <Action.ESE: -15>, <Action.NNE: 25>, <Action.SSW: -25>, <Action.WSW: -11>, <Action.SSW: -25>, <Action.SSW: -25>, <Action.SSW: -25>, <Action.WSW: -11>, <Action.WNW: 15>, <Action.NNE: 25>, <Action.NNE: 25>, <Action.NNE: 25>, <Action.WSW: -11>, <Action.WSW: -11>, <Action.NNE: 25>, <Action.SSW: -25>, <Action.NNE: 25>, <Action.ESE: -15>, <Action.NNE: 25>, <Action.NNW: 27>, <Action.WSW: -11>, <Action.WNW: 15>, <Action.SSE: -27>, <Action.NNE: 25>, <Action.NNE: 25>, <Action.ESE: -15>, <Action.WSW: -11>, <Action.WSW: -11>, <Action.WSW: -11>, <Action.NNE: 25>, <Action.SSW: -25>, <Action.WSW: -11>, <Action.ENE: 11>, <Action.NNE: 25>, <Action.SSW: -25>, <Action.NNW: 27>, <Action.SSW: -25>, <Action.SSW: -25>, <Action.WNW: 15>, <Action.ENE: 11>, <Action.NNE: 25>, <Action.WNW: 15>, <Action.NNE: 25>, <Action.SSE: -27>, <Action.NNE: 25>, <Action.NNE: 25>, <Action.WSW: -11>, <Action.ESE: -15>, <Action.SSW: -25>, <Action.ENE: 11>, <Action.SSE: -27>, <Action.ESE: -15>, <Action.ESE: -15>, <Action.SSW: -25>, <Action.WSW: -11>, <Action.SSE: -27>, <Action.NNE: 25>, <Action.SSW: -25>, <Action.WNW: 15>, <Action.WNW: 15>, <Action.NNE: 25>, <Action.NNE: 25>, <Action.NNE: 25>, <Action.ENE: 11>, <Action.WSW: -11>, <Action.SSW: -25>, <Action.NNE: 25>]\nState: Isolation(board=41523161203939122082683632224299007, ply_count=0, locs=(None, None)) \n Action: 108\n\n\nGot an int action: loc_sym = 110\nGot an int action: loc_sym = 4\nGot an int action: loc_sym = 6\nState: Isolation(board=41198642650280695355900476203722751, ply_count=1, locs=(108, None)) \n Action: 2\n\n\nGot an int action: loc_sym = 8\nGot an int action: loc_sym = 106\nGot an int action: loc_sym = 112\nState: Isolation(board=41198642650280695355900476203722747, ply_count=2, locs=(108, 2)) \n Action: -15\n\n\nState: Isolation(board=41198632746760381072858277010728955, ply_count=3, locs=(93, 2)) \n Action: 25\n\n\nResults for match 4: CustomPlayer2 wins.\n[58, 71, <Action.NNE: 25>, <Action.ENE: 11>, <Action.NNE: 25>, <Action.NNE: 25>, <Action.ESE: -15>, <Action.SSE: -27>, <Action.SSE: -27>, <Action.SSE: -27>, <Action.SSE: -27>, <Action.SSW: -25>, <Action.WNW: 15>, <Action.SSW: -25>, <Action.SSW: -25>, <Action.WNW: 15>, <Action.WNW: 15>, <Action.NNE: 25>, <Action.WNW: 15>, <Action.NNE: 25>, <Action.ENE: 11>, <Action.SSE: -27>, <Action.NNE: 25>, <Action.WSW: -11>, <Action.WSW: -11>, <Action.ESE: -15>, <Action.SSE: -27>, <Action.WSW: -11>, <Action.SSW: -25>, <Action.NNW: 27>, <Action.ESE: -15>, <Action.SSW: -25>, <Action.ESE: -15>, <Action.WNW: 15>, <Action.NNE: 25>, <Action.NNE: 25>, <Action.NNE: 25>, <Action.SSE: -27>, <Action.WNW: 15>, <Action.WNW: 15>, <Action.NNW: 27>, <Action.ENE: 11>, <Action.SSW: -25>, <Action.SSW: -25>, <Action.NNW: 27>, <Action.WNW: 15>, <Action.WSW: -11>, <Action.NNE: 25>, <Action.NNE: 25>, <Action.SSE: -27>, <Action.WSW: -11>, <Action.WNW: 15>, <Action.SSE: -27>, <Action.NNW: 27>, <Action.NNE: 25>, <Action.NNE: 25>, <Action.WSW: -11>, <Action.SSE: -27>, <Action.NNE: 25>]\nState: Isolation(board=41523161203939122082683632224299007, ply_count=0, locs=(None, None)) \n Action: 58\n\n\nGot an int action: loc_sym = 56\nGot an int action: loc_sym = 58\nGot an int action: loc_sym = 56\nState: Isolation(board=41523161203939121794453256072587263, ply_count=1, locs=(58, None)) \n Action: 71\n\n\nGot an int action: loc_sym = 69\nGot an int action: loc_sym = 45\nGot an int action: loc_sym = 43\nState: Isolation(board=41523161203936760611211821249980415, ply_count=2, locs=(58, 71)) \n Action: 25\n\n\nState: Isolation(board=41523161194265354054294787852331007, ply_count=3, locs=(83, 71)) \n Action: 11\n\n\n" ], [ "sum(abs(value) for value in book.values())", "_____no_output_____" ], [ "seconds = 29403 * 37 / 10\nprint('{} seconds'.format(seconds))\nprint('{} hours'.format(seconds/3600))", "108791.1 seconds\n30.21975 hours\n" ], [ "game_history", "_____no_output_____" ] ], [ [ "## Let's add the symmetry conditions to the game processing", "_____no_output_____" ] ], [ [ "s_a = list(book.keys())[0]\ns_a", "_____no_output_____" ], [ "W, H = 11, 9\n\ndef h_symmetry(loc):\n if loc is None:\n return None\n row = loc // (W + 2)\n center = W + (row - 1) * (W + 2) + (W + 2) // 2 + 1 if row != 0 else W // 2\n return 2 * center - loc", "_____no_output_____" ], [ "h_symmetry(28)", "2\n31\n" ], [ "h_symmetry(1)", "0\n5\n" ], [ "center = (H // 2) * (W + 2) + W // 2\ncenter", "_____no_output_____" ], [ "def c_symmetry(loc):\n if loc is None:\n return None\n center = (H // 2) * (W + 2) + W // 2\n return 2 * center - loc", "_____no_output_____" ], [ "c_symmetry(81)", "_____no_output_____" ], [ "c_symmetry(67)", "_____no_output_____" ], [ "def v_symmetry(loc):\n if loc is None:\n return None\n col = loc % (W + 2)\n center = (H // 2) * (W + 2) + col\n return 2 * center - loc", "_____no_output_____" ], [ "v_symmetry(2)", "_____no_output_____" ], [ "v_symmetry(28)", "_____no_output_____" ], [ "v_symmetry(48)", "_____no_output_____" ], [ "v_symmetry(86)", "_____no_output_____" ], [ "symmetric = b.sym_sa(s_a, loc_sym=h_symmetry, cardinal_sym=b.cardinal_sym_h)\nsymmetric", "8\n109\n6\n83\n" ], [ "print(isolation.DebugState.from_state(s_a[0]))", "\n+ - + - + - + - + - + - + - + - + - + - + - +\n| | | | | | | 1 | | | | |\n+ - + - + - + - + - + - + - + - + - + - + - +\n| | | | | | | | | X | | |\n+ - + - + - + - + - + - + - + - + - + - + - +\n| | | | | | 2 | | | | | |\n+ - + - + - + - + - + - + - + - + - + - + - +\n| | | | | | | | | | | |\n+ - + - + - + - + - + - + - + - + - + - + - +\n| | | | | | | | | | | |\n+ - + - + - + - + - + - + - + - + - + - + - +\n| | | | | | | | | | | |\n+ - + - + - + - + - + - + - + - + - + - + - +\n| | | | | | | | | | | |\n+ - + - + - + - + - + - + - + - + - + - + - +\n| | | | | | | | | | | |\n+ - + - + - + - + - + - + - + - + - + - + - +\n| | | | | | | | | | | |\n+ - + - + - + - + - + - + - + - + - + - + - +\n\n" ], [ "print(isolation.DebugState.from_state(symmetric[0]))", "\n+ - + - + - + - + - + - + - + - + - + - + - +\n| | | | | 1 | | | | | | |\n+ - + - + - + - + - + - + - + - + - + - + - +\n| | | | | | | | | | | |\n+ - + - + - + - + - + - + - + - + - + - + - +\n| | | | | | 2 | | | | | |\n+ - + - + - + - + - + - + - + - + - + - + - +\n| | | | | | | | | | | |\n+ - + - + - + - + - + - + - + - + - + - + - +\n| | | | | | | | | | | |\n+ - + - + - + - + - + - + - + - + - + - + - +\n| | | | | | | | | | | |\n+ - + - + - + - + - + - + - + - + - + - + - +\n| | | | | | | | | | | |\n+ - + - + - + - + - + - + - + - + - + - + - +\n| | | | | | | | | | | |\n+ - + - + - + - + - + - + - + - + - + - + - +\n| | | | | | | | | | | |\n+ - + - + - + - + - + - + - + - + - + - + - +\n\n" ], [ "def process_game_history(state,\n game_history,\n book,\n winner_id,\n active_player=0,\n depth=4):\n \"\"\" Given an initial state, and a list of actions, this function iterates\n through the resulting states of the actions and updates count of wins in\n the state/action book\"\"\"\n OPENING_MOVES = 2\n game_value = 2 * (active_player == winner_id) - 1\n curr_state = state # It is a named tuple, so I think it is immutable. No need to copy.\n for num_action, action in enumerate(game_history):\n if (curr_state, action) in book.keys():\n book[(curr_state, action)] += game_value\n if curr_state.ply_count <= OPENING_MOVES:\n book[b.sym_sa((curr_state, action), \n loc_sym=h_symmetry,\n cardinal_sym=b.cardinal_sym_h)] += game_value\n book[b.sym_sa((curr_state, action), \n loc_sym=v_symmetry,\n cardinal_sym=b.cardinal_sym_v)] += game_value\n book[b.sym_sa((curr_state, action), \n loc_sym=c_symmetry,\n cardinal_sym=b.cardinal_sym_c)] += game_value\n curr_state = curr_state.result(action)\n active_player = 1 - active_player\n game_value = 2 * (active_player == winner_id) - 1\n # Break on depth equal to book\n if num_action >= depth - 1:\n break", "_____no_output_____" ] ] ]

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

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

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "# Statistical Downscaling and Bias-Adjustment\n\n`xclim` provides tools and utilities to ease the bias-adjustement process through its `xclim.sdba` module. Almost all adjustment algorithms conform to the `train` - `adjust` scheme, formalized within `TrainAdjust` classes. Given a reference time series (ref), historical simulations (hist) and simulations to be adjusted (sim), any bias-adjustment method would be applied by first estimating the adjustment factors between the historical simulation and the observations series, and then applying these factors to `sim`, which could be a future simulation.\n\nThis presents examples, while a bit more info and the API are given on [this page](../sdba.rst).\n\nA very simple \"Quantile Mapping\" approach is available through the \"Empirical Quantile Mapping\" object. The object is created through the `.train` method of the class, and the simulation is adjusted with `.adjust`.", "_____no_output_____" ] ], [ [ "from __future__ import annotations\n\nimport cftime\nimport matplotlib.pyplot as plt\nimport numpy as np\nimport xarray as xr\n\n%matplotlib inline\nplt.style.use(\"seaborn\")\nplt.rcParams[\"figure.figsize\"] = (11, 5)\n\n# Create toy data to explore bias adjustment, here fake temperature timeseries\nt = xr.cftime_range(\"2000-01-01\", \"2030-12-31\", freq=\"D\", calendar=\"noleap\")\nref = xr.DataArray(\n (\n -20 * np.cos(2 * np.pi * t.dayofyear / 365)\n + 2 * np.random.random_sample((t.size,))\n + 273.15\n + 0.1 * (t - t[0]).days / 365\n ), # \"warming\" of 1K per decade,\n dims=(\"time\",),\n coords={\"time\": t},\n attrs={\"units\": \"K\"},\n)\nsim = xr.DataArray(\n (\n -18 * np.cos(2 * np.pi * t.dayofyear / 365)\n + 2 * np.random.random_sample((t.size,))\n + 273.15\n + 0.11 * (t - t[0]).days / 365\n ), # \"warming\" of 1.1K per decade\n dims=(\"time\",),\n coords={\"time\": t},\n attrs={\"units\": \"K\"},\n)\n\nref = ref.sel(time=slice(None, \"2015-01-01\"))\nhist = sim.sel(time=slice(None, \"2015-01-01\"))\n\nref.plot(label=\"Reference\")\nsim.plot(label=\"Model\")\nplt.legend()", "_____no_output_____" ], [ "from xclim import sdba\n\nQM = sdba.EmpiricalQuantileMapping.train(\n ref, hist, nquantiles=15, group=\"time\", kind=\"+\"\n)\nscen = QM.adjust(sim, extrapolation=\"constant\", interp=\"nearest\")\n\nref.groupby(\"time.dayofyear\").mean().plot(label=\"Reference\")\nhist.groupby(\"time.dayofyear\").mean().plot(label=\"Model - biased\")\nscen.sel(time=slice(\"2000\", \"2015\")).groupby(\"time.dayofyear\").mean().plot(\n label=\"Model - adjusted - 2000-15\", linestyle=\"--\"\n)\nscen.sel(time=slice(\"2015\", \"2030\")).groupby(\"time.dayofyear\").mean().plot(\n label=\"Model - adjusted - 2015-30\", linestyle=\"--\"\n)\nplt.legend()", "_____no_output_____" ] ], [ [ "In the previous example, a simple Quantile Mapping algorithm was used with 15 quantiles and one group of values. The model performs well, but our toy data is also quite smooth and well-behaved so this is not surprising. A more complex example could have biais distribution varying strongly across months. To perform the adjustment with different factors for each months, one can pass `group='time.month'`. Moreover, to reduce the risk of sharp change in the adjustment at the interface of the months, `interp='linear'` can be passed to `adjust` and the adjustment factors will be interpolated linearly. Ex: the factors for the 1st of May will be the average of those for april and those for may.", "_____no_output_____" ] ], [ [ "QM_mo = sdba.EmpiricalQuantileMapping.train(\n ref, hist, nquantiles=15, group=\"time.month\", kind=\"+\"\n)\nscen = QM_mo.adjust(sim, extrapolation=\"constant\", interp=\"linear\")\n\nref.groupby(\"time.dayofyear\").mean().plot(label=\"Reference\")\nhist.groupby(\"time.dayofyear\").mean().plot(label=\"Model - biased\")\nscen.sel(time=slice(\"2000\", \"2015\")).groupby(\"time.dayofyear\").mean().plot(\n label=\"Model - adjusted - 2000-15\", linestyle=\"--\"\n)\nscen.sel(time=slice(\"2015\", \"2030\")).groupby(\"time.dayofyear\").mean().plot(\n label=\"Model - adjusted - 2015-30\", linestyle=\"--\"\n)\nplt.legend()", "_____no_output_____" ] ], [ [ "The training data (here the adjustment factors) is available for inspection in the `ds` attribute of the adjustment object.", "_____no_output_____" ] ], [ [ "QM_mo.ds", "_____no_output_____" ], [ "QM_mo.ds.af.plot()", "_____no_output_____" ] ], [ [ "## Grouping\n\nFor basic time period grouping (months, day of year, season), passing a string to the methods needing it is sufficient. Most methods acting on grouped data also accept a `window` int argument to pad the groups with data from adjacent ones. Units of `window` are the sampling frequency of the main grouping dimension (usually `time`). For more complex grouping, or simply for clarity, one can pass a `xclim.sdba.base.Grouper` directly.\n\nExample here with another, simpler, adjustment method. Here we want `sim` to be scaled so that its mean fits the one of `ref`. Scaling factors are to be computed separately for each day of the year, but including 15 days on either side of the day. This means that the factor for the 1st of May is computed including all values from the 16th of April to the 15th of May (of all years).", "_____no_output_____" ] ], [ [ "group = sdba.Grouper(\"time.dayofyear\", window=31)\nQM_doy = sdba.Scaling.train(ref, hist, group=group, kind=\"+\")\nscen = QM_doy.adjust(sim)\n\nref.groupby(\"time.dayofyear\").mean().plot(label=\"Reference\")\nhist.groupby(\"time.dayofyear\").mean().plot(label=\"Model - biased\")\nscen.sel(time=slice(\"2000\", \"2015\")).groupby(\"time.dayofyear\").mean().plot(\n label=\"Model - adjusted - 2000-15\", linestyle=\"--\"\n)\nscen.sel(time=slice(\"2015\", \"2030\")).groupby(\"time.dayofyear\").mean().plot(\n label=\"Model - adjusted - 2015-30\", linestyle=\"--\"\n)\nplt.legend()", "_____no_output_____" ], [ "sim", "_____no_output_____" ], [ "QM_doy.ds.af.plot()", "_____no_output_____" ] ], [ [ "## Modular approach\n\nThe `sdba` module adopts a modular approach instead of implementing published and named methods directly.\nA generic bias adjustment process is laid out as follows:\n\n- preprocessing on `ref`, `hist` and `sim` (using methods in `xclim.sdba.processing` or `xclim.sdba.detrending`)\n- creating and training the adjustment object `Adj = Adjustment.train(obs, hist, **kwargs)` (from `xclim.sdba.adjustment`)\n- adjustment `scen = Adj.adjust(sim, **kwargs)`\n- post-processing on `scen` (for example: re-trending)\n\nThe train-adjust approach allows to inspect the trained adjustment object. The training information is stored in the underlying `Adj.ds` dataset and often has a `af` variable with the adjustment factors. Its layout and the other available variables vary between the different algorithm, refer to their part of the API docs.\n\nFor heavy processing, this separation allows the computation and writing to disk of the training dataset before performing the adjustment(s). See the [advanced notebook](sdba-advanced.ipynb).\n\nParameters needed by the training and the adjustment are saved to the `Adj.ds` dataset as a `adj_params` attribute. Other parameters, those only needed by the adjustment are passed in the `adjust` call and written to the history attribute in the output scenario dataarray.\n\n### First example : pr and frequency adaptation\n\nThe next example generates fake precipitation data and adjusts the `sim` timeseries but also adds a step where the dry-day frequency of `hist` is adapted so that is fits the one of `ref`. This ensures well-behaved adjustment factors for the smaller quantiles. Note also that we are passing `kind='*'` to use the multiplicative mode. Adjustment factors will be multiplied/divided instead of being added/substracted.", "_____no_output_____" ] ], [ [ "vals = np.random.randint(0, 1000, size=(t.size,)) / 100\nvals_ref = (4 ** np.where(vals < 9, vals / 100, vals)) / 3e6\nvals_sim = (\n (1 + 0.1 * np.random.random_sample((t.size,)))\n * (4 ** np.where(vals < 9.5, vals / 100, vals))\n / 3e6\n)\n\npr_ref = xr.DataArray(\n vals_ref, coords={\"time\": t}, dims=(\"time\",), attrs={\"units\": \"mm/day\"}\n)\npr_ref = pr_ref.sel(time=slice(\"2000\", \"2015\"))\npr_sim = xr.DataArray(\n vals_sim, coords={\"time\": t}, dims=(\"time\",), attrs={\"units\": \"mm/day\"}\n)\npr_hist = pr_sim.sel(time=slice(\"2000\", \"2015\"))\n\npr_ref.plot(alpha=0.9, label=\"Reference\")\npr_sim.plot(alpha=0.7, label=\"Model\")\nplt.legend()", "_____no_output_____" ], [ "# 1st try without adapt_freq\nQM = sdba.EmpiricalQuantileMapping.train(\n pr_ref, pr_hist, nquantiles=15, kind=\"*\", group=\"time\"\n)\nscen = QM.adjust(pr_sim)\n\npr_ref.sel(time=\"2010\").plot(alpha=0.9, label=\"Reference\")\npr_hist.sel(time=\"2010\").plot(alpha=0.7, label=\"Model - biased\")\nscen.sel(time=\"2010\").plot(alpha=0.6, label=\"Model - adjusted\")\nplt.legend()", "_____no_output_____" ] ], [ [ "In the figure above, `scen` has small peaks where `sim` is 0. This problem originates from the fact that there are more \"dry days\" (days with almost no precipitation) in `hist` than in `ref`. The next example works around the problem using frequency-adaptation, as described in [Themeßl et al. (2010)](https://doi.org/10.1007/s10584-011-0224-4).", "_____no_output_____" ] ], [ [ "# 2nd try with adapt_freq\nsim_ad, pth, dP0 = sdba.processing.adapt_freq(\n pr_ref, pr_sim, thresh=\"0.05 mm d-1\", group=\"time\"\n)\nQM_ad = sdba.EmpiricalQuantileMapping.train(\n pr_ref, sim_ad, nquantiles=15, kind=\"*\", group=\"time\"\n)\nscen_ad = QM_ad.adjust(pr_sim)\n\npr_ref.sel(time=\"2010\").plot(alpha=0.9, label=\"Reference\")\npr_sim.sel(time=\"2010\").plot(alpha=0.7, label=\"Model - biased\")\nscen_ad.sel(time=\"2010\").plot(alpha=0.6, label=\"Model - adjusted\")\nplt.legend()", "_____no_output_____" ] ], [ [ "### Second example: tas and detrending\n\nThe next example reuses the fake temperature timeseries generated at the beginning and applies the same QM adjustment method. However, for a better adjustment, we will scale sim to ref and then detrend the series, assuming the trend is linear. When `sim` (or `sim_scl`) is detrended, its values are now anomalies, so we need to normalize `ref` and `hist` so we can compare similar values.\n\nThis process is detailed here to show how the sdba module should be used in custom adjustment processes, but this specific method also exists as `sdba.DetrendedQuantileMapping` and is based on [Cannon et al. 2015](https://doi.org/10.1175/JCLI-D-14-00754.1). However, `DetrendedQuantileMapping` normalizes over a `time.dayofyear` group, regardless of what is passed in the `group` argument. As done here, it is anyway recommended to use `dayofyear` groups when normalizing, especially for variables with strong seasonal variations.", "_____no_output_____" ] ], [ [ "doy_win31 = sdba.Grouper(\"time.dayofyear\", window=15)\nSca = sdba.Scaling.train(ref, hist, group=doy_win31, kind=\"+\")\nsim_scl = Sca.adjust(sim)\n\ndetrender = sdba.detrending.PolyDetrend(degree=1, group=\"time.dayofyear\", kind=\"+\")\nsim_fit = detrender.fit(sim_scl)\nsim_detrended = sim_fit.detrend(sim_scl)\n\nref_n, _ = sdba.processing.normalize(ref, group=doy_win31, kind=\"+\")\nhist_n, _ = sdba.processing.normalize(hist, group=doy_win31, kind=\"+\")\n\nQM = sdba.EmpiricalQuantileMapping.train(\n ref_n, hist_n, nquantiles=15, group=\"time.month\", kind=\"+\"\n)\nscen_detrended = QM.adjust(sim_detrended, extrapolation=\"constant\", interp=\"nearest\")\nscen = sim_fit.retrend(scen_detrended)\n\n\nref.groupby(\"time.dayofyear\").mean().plot(label=\"Reference\")\nsim.groupby(\"time.dayofyear\").mean().plot(label=\"Model - biased\")\nscen.sel(time=slice(\"2000\", \"2015\")).groupby(\"time.dayofyear\").mean().plot(\n label=\"Model - adjusted - 2000-15\", linestyle=\"--\"\n)\nscen.sel(time=slice(\"2015\", \"2030\")).groupby(\"time.dayofyear\").mean().plot(\n label=\"Model - adjusted - 2015-30\", linestyle=\"--\"\n)\nplt.legend()", "_____no_output_____" ] ], [ [ "### Third example : Multi-method protocol - Hnilica et al. 2017\nIn [their paper of 2017](https://doi.org/10.1002/joc.4890), Hnilica, Hanel and Puš present a bias-adjustment method based on the principles of Principal Components Analysis. The idea is simple : use principal components to define coordinates on the reference and on the simulation and then transform the simulation data from the latter to the former. Spatial correlation can thus be conserved by taking different points as the dimensions of the transform space. The method was demonstrated in the article by bias-adjusting precipitation over different drainage basins.\n\nThe same method could be used for multivariate adjustment. The principle would be the same, concatening the different variables into a single dataset along a new dimension. An example is given in the [advanced notebook](sdba-advanced.ipynb).\n\nHere we show how the modularity of `xclim.sdba` can be used to construct a quite complex adjustment protocol involving two adjustment methods : quantile mapping and principal components. Evidently, as this example uses only 2 years of data, it is not complete. It is meant to show how the adjustment functions and how the API can be used.", "_____no_output_____" ] ], [ [ "# We are using xarray's \"air_temperature\" dataset\nds = xr.tutorial.open_dataset(\"air_temperature\")", "_____no_output_____" ], [ "# To get an exagerated example we select different points\n# here \"lon\" will be our dimension of two \"spatially correlated\" points\nreft = ds.air.isel(lat=21, lon=[40, 52]).drop_vars([\"lon\", \"lat\"])\nsimt = ds.air.isel(lat=18, lon=[17, 35]).drop_vars([\"lon\", \"lat\"])\n\n# Principal Components Adj, no grouping and use \"lon\" as the space dimensions\nPCA = sdba.PrincipalComponents.train(reft, simt, group=\"time\", crd_dim=\"lon\")\nscen1 = PCA.adjust(simt)\n\n# QM, no grouping, 20 quantiles and additive adjustment\nEQM = sdba.EmpiricalQuantileMapping.train(\n reft, scen1, group=\"time\", nquantiles=50, kind=\"+\"\n)\nscen2 = EQM.adjust(scen1)", "_____no_output_____" ], [ "# some Analysis figures\nfig = plt.figure(figsize=(12, 16))\ngs = plt.matplotlib.gridspec.GridSpec(3, 2, fig)\n\naxPCA = plt.subplot(gs[0, :])\naxPCA.scatter(reft.isel(lon=0), reft.isel(lon=1), s=20, label=\"Reference\")\naxPCA.scatter(simt.isel(lon=0), simt.isel(lon=1), s=10, label=\"Simulation\")\naxPCA.scatter(scen2.isel(lon=0), scen2.isel(lon=1), s=3, label=\"Adjusted - PCA+EQM\")\naxPCA.set_xlabel(\"Point 1\")\naxPCA.set_ylabel(\"Point 2\")\naxPCA.set_title(\"PC-space\")\naxPCA.legend()\n\nrefQ = reft.quantile(EQM.ds.quantiles, dim=\"time\")\nsimQ = simt.quantile(EQM.ds.quantiles, dim=\"time\")\nscen1Q = scen1.quantile(EQM.ds.quantiles, dim=\"time\")\nscen2Q = scen2.quantile(EQM.ds.quantiles, dim=\"time\")\nfor i in range(2):\n if i == 0:\n axQM = plt.subplot(gs[1, 0])\n else:\n axQM = plt.subplot(gs[1, 1], sharey=axQM)\n axQM.plot(refQ.isel(lon=i), simQ.isel(lon=i), label=\"No adj\")\n axQM.plot(refQ.isel(lon=i), scen1Q.isel(lon=i), label=\"PCA\")\n axQM.plot(refQ.isel(lon=i), scen2Q.isel(lon=i), label=\"PCA+EQM\")\n axQM.plot(\n refQ.isel(lon=i), refQ.isel(lon=i), color=\"k\", linestyle=\":\", label=\"Ideal\"\n )\n axQM.set_title(f\"QQ plot - Point {i + 1}\")\n axQM.set_xlabel(\"Reference\")\n axQM.set_xlabel(\"Model\")\n axQM.legend()\n\naxT = plt.subplot(gs[2, :])\nreft.isel(lon=0).plot(ax=axT, label=\"Reference\")\nsimt.isel(lon=0).plot(ax=axT, label=\"Unadjusted sim\")\n# scen1.isel(lon=0).plot(ax=axT, label='PCA only')\nscen2.isel(lon=0).plot(ax=axT, label=\"PCA+EQM\")\naxT.legend()\naxT.set_title(\"Timeseries - Point 1\")", "_____no_output_____" ] ], [ [ "### Fourth example : Multivariate bias-adjustment with multiple steps - Cannon 2018\n\nThis section replicates the \"MBCn\" algorithm described by [Cannon (2018)](https://doi.org/10.1007/s00382-017-3580-6). The method relies on some univariate algorithm, an adaption of the N-pdf transform of [Pitié et al. (2005)](https://ieeexplore.ieee.org/document/1544887/) and a final reordering step.\n\nIn the following, we use the AHCCD and CanESM2 data are reference and simulation and we correct both `pr` and `tasmax` together.", "_____no_output_____" ] ], [ [ "from xclim.core.units import convert_units_to\nfrom xclim.testing import open_dataset\n\ndref = open_dataset(\n \"sdba/ahccd_1950-2013.nc\", chunks={\"location\": 1}, drop_variables=[\"lat\", \"lon\"]\n).sel(time=slice(\"1981\", \"2010\"))\ndref = dref.assign(\n tasmax=convert_units_to(dref.tasmax, \"K\"),\n pr=convert_units_to(dref.pr, \"kg m-2 s-1\"),\n)\ndsim = open_dataset(\n \"sdba/CanESM2_1950-2100.nc\", chunks={\"location\": 1}, drop_variables=[\"lat\", \"lon\"]\n)\n\ndhist = dsim.sel(time=slice(\"1981\", \"2010\"))\ndsim = dsim.sel(time=slice(\"2041\", \"2070\"))\ndref", "_____no_output_____" ] ], [ [ "##### Perform an initial univariate adjustment.", "_____no_output_____" ] ], [ [ "# additive for tasmax\nQDMtx = sdba.QuantileDeltaMapping.train(\n dref.tasmax, dhist.tasmax, nquantiles=20, kind=\"+\", group=\"time\"\n)\n# Adjust both hist and sim, we'll feed both to the Npdf transform.\nscenh_tx = QDMtx.adjust(dhist.tasmax)\nscens_tx = QDMtx.adjust(dsim.tasmax)\n\n# remove == 0 values in pr:\ndref[\"pr\"] = sdba.processing.jitter_under_thresh(dref.pr, \"0.01 mm d-1\")\ndhist[\"pr\"] = sdba.processing.jitter_under_thresh(dhist.pr, \"0.01 mm d-1\")\ndsim[\"pr\"] = sdba.processing.jitter_under_thresh(dsim.pr, \"0.01 mm d-1\")\n\n# multiplicative for pr\nQDMpr = sdba.QuantileDeltaMapping.train(\n dref.pr, dhist.pr, nquantiles=20, kind=\"*\", group=\"time\"\n)\n# Adjust both hist and sim, we'll feed both to the Npdf transform.\nscenh_pr = QDMpr.adjust(dhist.pr)\nscens_pr = QDMpr.adjust(dsim.pr)\n\nscenh = xr.Dataset(dict(tasmax=scenh_tx, pr=scenh_pr))\nscens = xr.Dataset(dict(tasmax=scens_tx, pr=scens_pr))", "_____no_output_____" ] ], [ [ "##### Stack the variables to multivariate arrays and standardize them\nThe standardization process ensure the mean and standard deviation of each column (variable) is 0 and 1 respectively.\n\n`hist` and `sim` are standardized together so the two series are coherent. We keep the mean and standard deviation to be reused when we build the result.", "_____no_output_____" ] ], [ [ "# Stack the variables (tasmax and pr)\nref = sdba.processing.stack_variables(dref)\nscenh = sdba.processing.stack_variables(scenh)\nscens = sdba.processing.stack_variables(scens)\n\n# Standardize\nref, _, _ = sdba.processing.standardize(ref)\n\nallsim, savg, sstd = sdba.processing.standardize(xr.concat((scenh, scens), \"time\"))\nhist = allsim.sel(time=scenh.time)\nsim = allsim.sel(time=scens.time)", "_____no_output_____" ] ], [ [ "##### Perform the N-dimensional probability density function transform\n\nThe NpdfTransform will iteratively randomly rotate our arrays in the \"variables\" space and apply the univariate adjustment before rotating it back. In Cannon (2018) and Pitié et al. (2005), it can be seen that the source array's joint distribution converges toward the target's joint distribution when a large number of iterations is done.", "_____no_output_____" ] ], [ [ "from xclim import set_options\n\n# See the advanced notebook for details on how this option work\nwith set_options(sdba_extra_output=True):\n out = sdba.adjustment.NpdfTransform.adjust(\n ref,\n hist,\n sim,\n base=sdba.QuantileDeltaMapping, # Use QDM as the univariate adjustment.\n base_kws={\"nquantiles\": 20, \"group\": \"time\"},\n n_iter=20, # perform 20 iteration\n n_escore=1000, # only send 1000 points to the escore metric (it is realy slow)\n )\n\nscenh = out.scenh.rename(time_hist=\"time\") # Bias-adjusted historical period\nscens = out.scen # Bias-adjusted future period\nextra = out.drop_vars([\"scenh\", \"scen\"])\n\n# Un-standardize (add the mean and the std back)\nscenh = sdba.processing.unstandardize(scenh, savg, sstd)\nscens = sdba.processing.unstandardize(scens, savg, sstd)", "_____no_output_____" ] ], [ [ "##### Restoring the trend\n\nThe NpdfT has given us new \"hist\" and \"sim\" arrays with a correct rank structure. However, the trend is lost in this process. We reorder the result of the initial adjustment according to the rank structure of the NpdfT outputs to get our final bias-adjusted series.\n\n`sdba.processing.reordering` : 'ref' the argument that provides the order, 'sim' is the argument to reorder.", "_____no_output_____" ] ], [ [ "scenh = sdba.processing.reordering(hist, scenh, group=\"time\")\nscens = sdba.processing.reordering(sim, scens, group=\"time\")", "_____no_output_____" ], [ "scenh = sdba.processing.unstack_variables(scenh)\nscens = sdba.processing.unstack_variables(scens)", "_____no_output_____" ] ], [ [ "##### There we are!\n\nLet's trigger all the computations. Here we write the data to disk and use `compute=False` in order to trigger the whole computation tree only once. There seems to be no way in xarray to do the same with a `load` call.", "_____no_output_____" ] ], [ [ "from dask import compute\nfrom dask.diagnostics import ProgressBar\n\ntasks = [\n scenh.isel(location=2).to_netcdf(\"mbcn_scen_hist_loc2.nc\", compute=False),\n scens.isel(location=2).to_netcdf(\"mbcn_scen_sim_loc2.nc\", compute=False),\n extra.escores.isel(location=2)\n .to_dataset()\n .to_netcdf(\"mbcn_escores_loc2.nc\", compute=False),\n]\n\nwith ProgressBar():\n compute(tasks)", "_____no_output_____" ] ], [ [ "Let's compare the series and look at the distance scores to see how well the Npdf transform has converged.", "_____no_output_____" ] ], [ [ "scenh = xr.open_dataset(\"mbcn_scen_hist_loc2.nc\")\n\nfig, ax = plt.subplots()\n\ndref.isel(location=2).tasmax.plot(ax=ax, label=\"Reference\")\nscenh.tasmax.plot(ax=ax, label=\"Adjusted\", alpha=0.65)\ndhist.isel(location=2).tasmax.plot(ax=ax, label=\"Simulated\")\n\nax.legend()", "_____no_output_____" ], [ "escores = xr.open_dataarray(\"mbcn_escores_loc2.nc\")\ndiff_escore = escores.differentiate(\"iterations\")\ndiff_escore.plot()\nplt.title(\"Difference of the subsequent e-scores.\")\nplt.ylabel(\"E-scores difference\")", "_____no_output_____" ], [ "diff_escore", "_____no_output_____" ] ], [ [ "The tutorial continues in the [advanced notebook](sdba-advanced.ipynb) with more on optimization with dask, other fancier detrending algorithms and an example pipeline for heavy processing.", "_____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" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Working with PDBsum in Jupyter & Demonstration of PDBsum protein interface data to dataframe script", "_____no_output_____" ], [ "Usually you'll want to get some data from PDBsum and analyze it. For the current example in this series of notebooks, I'll cover how to bring in a file of protein-protein interactions and then progress through using that in combination with Python to analyze the results and ultimately compare the results to a different structure.\n\n-----\n\n<div class=\"alert alert-block alert-warning\">\n<p>If you haven't used one of these notebooks before, they're basically web pages in which you can write, edit, and run live code. They're meant to encourage experimentation, so don't feel nervous. Just try running a few cells and see what happens!.</p>\n\n<p>\n Some tips:\n <ul>\n <li>Code cells have boxes around them. When you hover over them an <i class=\"fa-step-forward fa\"></i> icon appears.</li>\n <li>To run a code cell either click the <i class=\"fa-step-forward fa\"></i> icon, or click on the cell and then hit <b>Shift+Enter</b>. The <b>Shift+Enter</b> combo will also move you to the next cell, so it's a quick way to work through the notebook.</li>\n <li>While a cell is running a <b>*</b> appears in the square brackets next to the cell. Once the cell has finished running the asterisk will be replaced with a number.</li>\n <li>In most cases you'll want to start from the top of notebook and work your way down running each cell in turn. Later cells might depend on the results of earlier ones.</li>\n <li>To edit a code cell, just click on it and type stuff. Remember to run the cell once you've finished editing.</li>\n </ul>\n</p>\n</div>\n\n----", "_____no_output_____" ], [ "### Retrieving Protein-Protein interface reports/ the list of interactions\n\n#### Getting list of interactions between two proteins under individual entries under PDBsum's 'Prot-prot' tab via command line.\n\nSay example from [here](http://www.ebi.ac.uk/thornton-srv/databases/cgi-bin/pdbsum/GetPage.pl?pdbcode=6ah3&template=interfaces.html&o=RESIDUE&l=3) links to the following as 'List of\ninteractions' in the bottom right of the page:\n\n```text \nhttp://www.ebi.ac.uk/thornton-srv/databases/cgi-bin/pdbsum/GetIface.pl?pdb=6ah3&chain1=B&chain2=G\n```\n \nThen based on suggestion at top [here](https://stackoverflow.com/a/52363117/8508004) that would be used in a curl command where the items after the `?` in the original URL get placed into quotes and provided following the `--data` flag argument option in the call to `curl`, like so:\n```text\ncurl -L -o data.txt --data \"pdb=6ah3&chain1=B&chain2=G\" http://www.ebi.ac.uk/thornton-srv/databases/cgi-bin/pdbsum/GetIface.pl\n```\n\n**Specifically**, the `--data \"pdb=6ah3&chain1=B&chain2=G\"` is the part coming from the end of the original URL.\n\n\nPutting that into action in Jupyter to fetch for the example the interactions list in a text:", "_____no_output_____" ] ], [ [ "!curl -L -o data.txt --data \"pdb=6ah3&chain1=B&chain2=G\" http://www.ebi.ac.uk/thornton-srv/databases/cgi-bin/pdbsum/GetIface.pl", " % Total % Received % Xferd Average Speed Time Time Time Current\n Dload Upload Total Spent Left Speed\n100 7063 0 7037 100 26 9033 33 --:--:-- --:--:-- --:--:-- 9055\n" ] ], [ [ "To prove that the data file has been retieved, we'll show the first 16 lines of it by running the next cell:", "_____no_output_____" ] ], [ [ "!head -16 data.txt", "<PRE>\r\nList of atom-atom interactions across protein-protein interface\r\n---------------------------------------------------------------\r\n<P>\r\n PDB code: 6ah3 Chains B }{ G\r\n ------------------------------\r\n<P>\r\n\r\nHydrogen bonds\r\n--------------\r\n\r\n <----- A T O M 1 -----> <----- A T O M 2 ----->\r\n\r\n Atom Atom Res Res Atom Atom Res Res\r\n no. name name no. Chain no. name name no. Chain Distance\r\n 1. 9937 NZ LYS 326 B <--> 20598 O LYS 122 G 2.47\r\n" ] ], [ [ "Later in this series of notebooks, I'll demonstrate how to make this step even easier with just the PDB entry id and the chains you are interested in and the later how to loop on this process to get multiple data files for interactions from different structures.", "_____no_output_____" ], [ "### Making a Pandas dataframe from the interactions file\n\nTo convert the data to a dataframe, we'll use a script.\n\n If you haven't encountered Pandas dataframes before I suggest you see the first two notebooks that come up with you launch a session from my [blast-binder](https://github.com/fomightez/blast-binder) site. Those first two notebooks cover using the dataframe containing BLAST results some. \n \nTo get that script, you can run the next cell. (It is not included in the repository where this launches from to insure you always get the most current version, which is assumed to be the best available at the time.)", "_____no_output_____" ] ], [ [ "!curl -OL https://raw.githubusercontent.com/fomightez/structurework/master/pdbsum-utilities/pdbsum_prot_interactions_list_to_df.py", " % Total % Received % Xferd Average Speed Time Time Time Current\n Dload Upload Total Spent Left Speed\n100 23915 100 23915 0 0 35272 0 --:--:-- --:--:-- --:--:-- 35220\n" ] ], [ [ "We have the script now. And we already have a data file for it to process. To process the data file, run the next command where we use Python to run the script and direct it at the results file, `data.txt`, we made just a few cells ago.", "_____no_output_____" ] ], [ [ "%run pdbsum_prot_interactions_list_to_df.py data.txt", "Provided interactions data read and converted to a dataframe...\n\nA dataframe of the data has been saved as a file\nin a manner where other Python programs can access it (pickled form).\nRESULTING DATAFRAME is stored as ==> 'prot_int_pickled_df.pkl'" ] ], [ [ "As of writing this, the script we are using outputs a file that is a binary, compact form of the dataframe. (That means it is tiny and not human readable. It is called 'pickled'. Saving in that form may seem odd, but as illustrated [here](#Output-to-more-universal,-table-like-formats) below this is is a very malleable form. And even more pertinent for dealing with data in Jupyter notebooks, there is actually an easier way to interact with this script when in Jupyter notebook that skips saving this intermediate file. So hang on through the long, more trandtional way of doing this before the easier way is introduced. And I saved it in the compact form and not the mroe typical tab-delimited form because we mostly won't go this route and might as well make tiny files while working along to a better route. It is easy to convert back and forth using the pickled form assuming you can match the Pandas/Python versions.)\n\nWe can take that file where the dataframe is pickled, and bring it into active memory in this notebook with another command form the Pandas library. First, we have to import the Pandas library.\nRun the next command to bring the dataframe into active memory. Note the name comes from the name noted when we ran the script in the cell above.", "_____no_output_____" ] ], [ [ "import pandas as pd\ndf = pd.read_pickle(\"prot_int_pickled_df.pkl\")", "_____no_output_____" ] ], [ [ "When that last cell ran, you won't notice any output, but something happened. We can look at that dataframe by calling it in a cell.", "_____no_output_____" ] ], [ [ "df", "_____no_output_____" ] ], [ [ "You'll notice that if the list of data is large, that the Jupyter environment represents just the head and tail to make it more reasonable. There are ways you can have Jupyter display it all which we won't go into here. \n\nInstead we'll start to show some methods of dataframes that make them convenient. For example, you can use the `head` method to see the start like we used on the command line above.", "_____no_output_____" ] ], [ [ "df.head()", "_____no_output_____" ] ], [ [ "Now what types of interactions are observed for this pair of interacting protein chains?\n\nTo help answer that, we can group the results by the type column.", "_____no_output_____" ] ], [ [ "grouped = df.groupby('type')\nfor type, grouped_df in grouped:\n print(type)\n display(grouped_df)", "Hydrogen bonds\n" ] ], [ [ "Same data as earlier but we can cleary see we have Hydrogen bonds, Non-bonded contacts (a.k.a., van der Waals contacts), and salt bridges, and we immediately get a sense of what types of interactions are more abundant.\n\nYou may want to get a sense of what else you can do by examining he first two notebooks that come up with you launch a session from my [blast-binder](https://github.com/fomightez/blast-binder) site. Those first two notebooks cover using the dataframe containing BLAST results some.\n\nShortly, we'll cover how to bring the dataframe we just made into the notebook without dealing with a file intermediate; however, next I'll demonstrate how to save it as text for use elsewhere, such as in Excel.", "_____no_output_____" ], [ "## Output to more universal, table-like formats\n\nI've tried to sell you on the power of the Python/Pandas dataframe, but it isn't for all uses or everyone. However, most everyone is accustomed to dealing with text based tables or even Excel. In fact, a text-based based table perhaps tab or comma-delimited would be the better way to archive the data we are generating here. Python/Pandas makes it easy to go from the dataframe form to these tabular forms. You can even go back later from the table to the dataframe, which may be inportant if you are going to different versions of Python/Pandas as I briefly mentioned parenthetically above.\n\n**First, generating a text-based table.**", "_____no_output_____" ] ], [ [ "#Save / write a TSV-formatted (tab-separated values/ tab-delimited) file\ndf.to_csv('pdbsum_data.tsv', sep='\\t',index = False) #add `,header=False` to leave off header, too", "_____no_output_____" ] ], [ [ "Because `df.to_csv()` defaults to dealing with csv, you can simply use `df.to_csv('example.csv',index = False)` for comma-delimited (comma-separated) files.\n\nYou can see that worked by looking at the first few lines with the next command. (Feel free to make the number higher or delete the number all together. I restricted it just to first line to make output smaller.)", "_____no_output_____" ] ], [ [ "!head -5 pdbsum_data.tsv", "Atom1 no.\tAtom1 name\tAtom1 Res name\tAtom1 Res no.\tAtom1 Chain\tAtom2 no.\tAtom2 name\tAtom2 Res name\tAtom2 Res no.\tAtom2 Chain\tDistance\ttype\r\n9937\tNZ\tLYS\t326\tB\t20598\tO\tLYS\t122\tG\t2.47\tHydrogen bonds\r\n9591\tO\tCYS\t280\tB\t19928\tCG1\tILE\t29\tG\t3.77\tNon-bonded contacts\r\n9591\tO\tCYS\t280\tB\t19930\tCD1\tILE\t29\tG\t3.42\tNon-bonded contacts\r\n9593\tSG\tCYS\t280\tB\t19872\tNZ\tLYS\t22\tG\t3.81\tNon-bonded contacts\r\n" ] ], [ [ "If you had need to go back from a tab-separated table to a dataframe, you can run something like in the following cell.", "_____no_output_____" ] ], [ [ "reverted_df = pd.read_csv('pdbsum_data.tsv', sep='\\t')\nreverted_df.to_pickle('reverted_df.pkl') # OPTIONAL: pickle that data too", "_____no_output_____" ] ], [ [ "For a comma-delimited (CSV) file you'd use `df = pd.read_csv('example.csv')` because `pd.read_csv()` method defaults to comma as the separator (`sep` parameter).\n\nYou can verify that read from the text-based table by viewing it with the next line.", "_____no_output_____" ] ], [ [ "reverted_df.head()", "_____no_output_____" ] ], [ [ "**Generating an Excel spreadsheet from a dataframe.**\n\nBecause this is an specialized need, there is a special module needed that I didn't bother installing by default and so it needs to be installed before generating the Excel file. Running the next cell will do both.", "_____no_output_____" ] ], [ [ "%pip install openpyxl\n# save to excel (KEEPS multiINDEX, and makes sparse to look good in Excel straight out of Python)\ndf.to_excel('pdbsum_data.xlsx') # after openpyxl installed", "Requirement already satisfied: openpyxl in /srv/conda/envs/notebook/lib/python3.7/site-packages (3.0.6)\nRequirement already satisfied: et-xmlfile in /srv/conda/envs/notebook/lib/python3.7/site-packages (from openpyxl) (1.0.1)\nRequirement already satisfied: jdcal in /srv/conda/envs/notebook/lib/python3.7/site-packages (from openpyxl) (1.4.1)\nNote: you may need to restart the kernel to use updated packages.\n" ] ], [ [ "You'll need to download the file first to your computer and then view it locally as there is no viewer in the Jupyter environment.\n\nAdiitionally, it is possible to add styles to dataframes and the styles such as shading of cells and coloring of text will be translated to the Excel document made as well.\n\nExcel files can be read in to Pandas dataframes directly without needing to go to a text based intermediate first.", "_____no_output_____" ] ], [ [ "# read Excel\ndf_from_excel = pd.read_excel('pdbsum_data.xlsx',engine='openpyxl') # see https://stackoverflow.com/a/65266270/8508004 where notes xlrd no longer supports xlsx", "Collecting xlrd\n Downloading xlrd-2.0.1-py2.py3-none-any.whl (96 kB)\n\u001b[K |████████████████████████████████| 96 kB 2.8 MB/s eta 0:00:011\n\u001b[?25hInstalling collected packages: xlrd\nSuccessfully installed xlrd-2.0.1\nNote: you may need to restart the kernel to use updated packages.\n" ] ], [ [ "That can be viewed to convince yourself it worked by running the next command.", "_____no_output_____" ] ], [ [ "df_from_excel.head()", "_____no_output_____" ] ], [ [ "Next, we'll cover how to bring the dataframe we just made into the notebook without dealing with a file intermediate.\n\n----\n\n### Making a Pandas dataframe from the interactions file directly in Jupyter\n\nFirst we'll check for the script we'll use and get it if we don't already have it. \n\n(The thinking is once you know what you are doing you may have skipped all the steps above and not have the script you'll need yet. It cannot hurt to check and if it isn't present, bring it here.)", "_____no_output_____" ] ], [ [ "# Get a file if not yet retrieved / check if file exists\nimport os\nfile_needed = \"pdbsum_prot_interactions_list_to_df.py\"\nif not os.path.isfile(file_needed):\n !curl -OL https://raw.githubusercontent.com/fomightez/structurework/master/pdbsum-utilities/{file_needed}", "_____no_output_____" ] ], [ [ "This is going to rely on approaches very similar to those illustrated [here](https://github.com/fomightez/patmatch-binder/blob/6f7630b2ee061079a72cd117127328fd1abfa6c7/notebooks/PatMatch%20with%20more%20Python.ipynb#Passing-results-data-into-active-memory-without-a-file-intermediate) and [here](https://github.com/fomightez/patmatch-binder/blob/6f7630b2ee061079a72cd117127328fd1abfa6c7/notebooks/Sending%20PatMatch%20output%20directly%20to%20Python.ipynb##Running-Patmatch-and-passing-the-results-to-Python-without-creating-an-output-file-intermediate).\n\nWe obtained the `pdbsum_prot_interactions_list_to_df.py` script in the preparation steps above. However, instead of using it as an external script as we did earlier in this notebook, we want to use the core function of that script within this notebook for the options that involve no pickled-object file intermediate. Similar to the way we imported a lot of other useful modules in the first notebook and a cell above, you can run the next cell to bring in to memory of this notebook's computational environment, the main function associated with the `pdbsum_prot_interactions_list_to_df.py` script, aptly named `pdbsum_prot_interactions_list_to_df`. (As written below the command to do that looks a bit redundant;however, the first from part of the command below actually is referencing the `pdbsum_prot_interactions_list_to_df.py` script, but it doesn't need the `.py` extension because the import only deals with such files.)", "_____no_output_____" ] ], [ [ "from pdbsum_prot_interactions_list_to_df import pdbsum_prot_interactions_list_to_df", "_____no_output_____" ] ], [ [ "We can demonstrate that worked by calling the function.", "_____no_output_____" ] ], [ [ "pdbsum_prot_interactions_list_to_df()", "_____no_output_____" ] ], [ [ "If the module was not imported, you'd see `ModuleNotFoundError: No module named 'pdbsum_prot_interactions_list_to_df'`, but instead you should see it saying it is missing `data_file` to act on because you passed it nothing.\n\nAfter importing the main function of that script into this running notebook, you are ready to demonstrate the approach that doesn't require a file intermediates. The imported `pdbsum_prot_interactions_list_to_df` function is used within the computational environment of the notebook and the dataframe produced assigned to a variable in the running the notebook. In the end, the results are in an active dataframe in the notebook without needing to read the pickled dataframe. **Although bear in mind the pickled dataframe still gets made, and it is good to download and keep that pickled dataframe since you'll find it convenient for reading and getting back into an analysis without need for rerunning earlier steps again.**", "_____no_output_____" ] ], [ [ "direct_df = pdbsum_prot_interactions_list_to_df(\"data.txt\")\ndirect_df.head()", "Provided interactions data read and converted to a dataframe...\n\nA dataframe of the data has been saved as a file\nin a manner where other Python programs can access it (pickled form).\nRESULTING DATAFRAME is stored as ==> 'prot_int_pickled_df.pkl'\n\nReturning a dataframe with the information as well." ] ], [ [ "This may be how you prefer to use the script. Either option exists.\n\n----\n\nContinue on with the next notebook in the series, [Using PDBsum data to highlight changes in protein-protein interactions](Using%20PDBsum%20data%20to%20highlight%20changes%20in%20protein-protein%20interactions.ipynb). That notebook builds on the ground work here to demonstrate how to examine similarities and differences in specific residue-level interactions between the same chains in different, related structures.\n\n----", "_____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" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "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" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "empty" ] ] ]

[ "empty" ]

[ [ "empty" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "# !python -m pip install seaborn", "_____no_output_____" ], [ "# %load_ext autoreload\n# %autoreload 2", "_____no_output_____" ], [ "import seaborn as sns", "_____no_output_____" ], [ "df = sns.load_dataset('titanic')\ndf.shape", "_____no_output_____" ], [ "df.info()", "<class 'pandas.core.frame.DataFrame'>\nRangeIndex: 891 entries, 0 to 890\nData columns (total 15 columns):\n # Column Non-Null Count Dtype \n--- ------ -------------- ----- \n 0 survived 891 non-null int64 \n 1 pclass 891 non-null int64 \n 2 sex 891 non-null object \n 3 age 714 non-null float64 \n 4 sibsp 891 non-null int64 \n 5 parch 891 non-null int64 \n 6 fare 891 non-null float64 \n 7 embarked 889 non-null object \n 8 class 891 non-null category\n 9 who 891 non-null object \n 10 adult_male 891 non-null bool \n 11 deck 203 non-null category\n 12 embark_town 889 non-null object \n 13 alive 891 non-null object \n 14 alone 891 non-null bool \ndtypes: bool(2), category(2), float64(2), int64(4), object(5)\nmemory usage: 80.6+ KB\n" ] ], [ [ "survived, pclass, sibsp, parch, fare", "_____no_output_____" ] ], [ [ "X = df[['pclass', 'sibsp', 'parch', 'fare']]\nY = df[['survived']]\nX.shape, Y.shape", "_____no_output_____" ], [ "from sklearn.model_selection import train_test_split", "_____no_output_____" ], [ "x_train, x_test, y_train, y_test = train_test_split(X, Y)\nx_train.shape, x_test.shape, y_train.shape, y_test.shape", "_____no_output_____" ], [ "from sklearn.linear_model import LogisticRegression", "_____no_output_____" ], [ "logR = LogisticRegression()\ntype(logR)", "_____no_output_____" ], [ "logR.fit(x_train, y_train)", "/usr/local/lib/python3.7/dist-packages/sklearn/utils/validation.py:760: DataConversionWarning: A column-vector y was passed when a 1d array was expected. Please change the shape of y to (n_samples, ), for example using ravel().\n y = column_or_1d(y, warn=True)\n" ], [ "logR.classes_", "_____no_output_____" ], [ "logR.coef_\n# 'pclass', 'sibsp', 'parch', 'fare'", "_____no_output_____" ], [ "logR.score(x_train, y_train)", "_____no_output_____" ], [ "logR.predict(x_train)", "_____no_output_____" ], [ "logR.predict_proba(x_train)", "_____no_output_____" ], [ "logR.predict_proba(x_train[10:13])", "_____no_output_____" ], [ "0.41873577+0.58126423", "_____no_output_____" ], [ "logR.predict(x_train[10:13])", "_____no_output_____" ], [ "print('Hello')", "Hello\n" ], [ "from sklearn import metrics", "_____no_output_____" ], [ "metrics.confusion_matrix(x_train, y_train)", "_____no_output_____" ], [ "", "_____no_output_____" ] ] ]

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

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

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ [ [ "# Working with SeqFish data", "_____no_output_____" ] ], [ [ "import stlearn as st", "_____no_output_____" ] ], [ [ "The data is downloaded from https://www.spatialomics.org/SpatialDB/download.php\n\n| Technique | PMID | Title | Expression | SV genes|\n| ----------- | ----------- | ----------- | ----------- | ----------- |\n|seqFISH|30911168|Transcriptome-scale super-resolved imaging in tissues by RNA seqFISH+\tseqfish_30911168.tar.gz|seqfish_30911168_SVG.tar.gz\n\nRead SeqFish data and we select field 5.", "_____no_output_____" ] ], [ [ "data = st.ReadSeqFish(count_matrix_file=\"../Downloads/seqfish_30911168/cortex_svz_counts.matrix\",\n spatial_file=\"../Downloads/seqfish_30911168/cortex_svz_cellcentroids.csv\",\n field=5)", "D:\\Anaconda3\\envs\\test2\\lib\\site-packages\\anndata-0.7.3-py3.8.egg\\anndata\\_core\\anndata.py:119: ImplicitModificationWarning: Transforming to str index.\n warnings.warn(\"Transforming to str index.\", ImplicitModificationWarning)\n" ] ], [ [ "Quality checking for the data", "_____no_output_____" ] ], [ [ "st.pl.QC_plot(data)", "_____no_output_____" ] ], [ [ "Plot gene Nr4a1", "_____no_output_____" ] ], [ [ "st.pl.gene_plot(data,genes=\"Nr4a1\")", "_____no_output_____" ] ], [ [ "Running Preprocessing for MERFISH data", "_____no_output_____" ] ], [ [ "st.pp.filter_genes(data,min_cells=3)\nst.pp.normalize_total(data)\nst.pp.log1p(data)\nst.pp.scale(data)", "Normalization step is finished in adata.X\nLog transformation step is finished in adata.X\nScale step is finished in adata.X\n" ] ], [ [ "Running PCA to reduce the dimensions to 50", "_____no_output_____" ] ], [ [ "st.em.run_pca(data,n_comps=50,random_state=0)", "PCA is done! Generated in adata.obsm['X_pca'], adata.uns['pca'] and adata.varm['PCs']\n" ] ], [ [ "Perform Louvain clustering", "_____no_output_____" ] ], [ [ "st.pp.neighbors(data,n_neighbors=25)", "D:\\Anaconda3\\envs\\test2\\lib\\site-packages\\umap_learn-0.4.3-py3.8.egg\\umap\\spectral.py:4: NumbaDeprecationWarning: No direct replacement for 'numba.targets' available. Visit https://gitter.im/numba/numba-dev to request help. Thanks!\n import numba.targets\n" ], [ "st.tl.clustering.louvain(data)", "Applying Louvain clustering ...\nLouvain clustering is done! The labels are stored in adata.obs['louvain']\n" ], [ "st.pl.cluster_plot(data,use_label=\"louvain\",spot_size=10)", "_____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", "code" ] ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ [ [ "# Machine Learning Pipeline - Feature Engineering\n\nIn the following notebooks, we will go through the implementation of each one of the steps in the Machine Learning Pipeline. \n\nWe will discuss:\n\n1. Data Analysis\n2. **Feature Engineering**\n3. Feature Selection\n4. Model Training\n5. Obtaining Predictions / Scoring\n\n\nWe will use the house price dataset available on [Kaggle.com](https://www.kaggle.com/c/house-prices-advanced-regression-techniques/data). See below for more details.\n\n===================================================================================================\n\n## Predicting Sale Price of Houses\n\nThe aim of the project is to build a machine learning model to predict the sale price of homes based on different explanatory variables describing aspects of residential houses.\n\n\n### Why is this important? \n\nPredicting house prices is useful to identify fruitful investments, or to determine whether the price advertised for a house is over or under-estimated.\n\n\n### What is the objective of the machine learning model?\n\nWe aim to minimise the difference between the real price and the price estimated by our model. We will evaluate model performance with the:\n\n1. mean squared error (mse)\n2. root squared of the mean squared error (rmse)\n3. r-squared (r2).\n\n\n### How do I download the dataset?\n\n- Visit the [Kaggle Website](https://www.kaggle.com/c/house-prices-advanced-regression-techniques/data).\n\n- Remember to **log in**\n\n- Scroll down to the bottom of the page, and click on the link **'train.csv'**, and then click the 'download' blue button towards the right of the screen, to download the dataset.\n\n- The download the file called **'test.csv'** and save it in the directory with the notebooks.\n\n\n**Note the following:**\n\n- You need to be logged in to Kaggle in order to download the datasets.\n- You need to accept the terms and conditions of the competition to download the dataset\n- If you save the file to the directory with the jupyter notebook, then you can run the code as it is written here.", "_____no_output_____" ], [ "# Reproducibility: Setting the seed\n\nWith the aim to ensure reproducibility between runs of the same notebook, but also between the research and production environment, for each step that includes some element of randomness, it is extremely important that we **set the seed**.", "_____no_output_____" ] ], [ [ "# to handle datasets\nimport pandas as pd\nimport numpy as np\n\n# for plotting\nimport matplotlib.pyplot as plt\n\n# for the yeo-johnson transformation\nimport scipy.stats as stats\n\n# to divide train and test set\nfrom sklearn.model_selection import train_test_split\n\n# feature scaling\nfrom sklearn.preprocessing import MinMaxScaler\n\n# to save the trained scaler class\nimport joblib\n\n# to visualise al the columns in the dataframe\npd.pandas.set_option('display.max_columns', None)", "_____no_output_____" ], [ "# load dataset\ndata = pd.read_csv('train.csv')\n\n# rows and columns of the data\nprint(data.shape)\n\n# visualise the dataset\ndata.head()", "(1460, 81)\n" ] ], [ [ "# Separate dataset into train and test\n\nIt is important to separate our data intro training and testing set. \n\nWhen we engineer features, some techniques learn parameters from data. It is important to learn these parameters only from the train set. This is to avoid over-fitting.\n\nOur feature engineering techniques will learn:\n\n- mean\n- mode\n- exponents for the yeo-johnson\n- category frequency\n- and category to number mappings\n\nfrom the train set.\n\n**Separating the data into train and test involves randomness, therefore, we need to set the seed.**", "_____no_output_____" ] ], [ [ "# Let's separate into train and test set\n# Remember to set the seed (random_state for this sklearn function)\n\nX_train, X_test, y_train, y_test = train_test_split(\n data.drop(['Id', 'SalePrice'], axis=1), # predictive variables\n data['SalePrice'], # target\n test_size=0.1, # portion of dataset to allocate to test set\n random_state=0, # we are setting the seed here\n)\n\nX_train.shape, X_test.shape", "_____no_output_____" ] ], [ [ "# Feature Engineering\n\nIn the following cells, we will engineer the variables of the House Price Dataset so that we tackle:\n\n1. Missing values\n2. Temporal variables\n3. Non-Gaussian distributed variables\n4. Categorical variables: remove rare labels\n5. Categorical variables: convert strings to numbers\n5. Put the variables in a similar scale", "_____no_output_____" ], [ "## Target\n\nWe apply the logarithm", "_____no_output_____" ] ], [ [ "y_train = np.log(y_train)\ny_test = np.log(y_test)", "_____no_output_____" ] ], [ [ "## Missing values\n\n### Categorical variables\n\nWe will replace missing values with the string \"missing\" in those variables with a lot of missing data. \n\nAlternatively, we will replace missing data with the most frequent category in those variables that contain fewer observations without values. \n\nThis is common practice.", "_____no_output_____" ] ], [ [ "# let's identify the categorical variables\n# we will capture those of type object\n\ncat_vars = [var for var in data.columns if data[var].dtype == 'O']\n\n# MSSubClass is also categorical by definition, despite its numeric values\n# (you can find the definitions of the variables in the data_description.txt\n# file available on Kaggle, in the same website where you downloaded the data)\n\n# lets add MSSubClass to the list of categorical variables\ncat_vars = cat_vars + ['MSSubClass']\n\n# cast all variables as categorical\nX_train[cat_vars] = X_train[cat_vars].astype('O')\nX_test[cat_vars] = X_test[cat_vars].astype('O')\n\n# number of categorical variables\nlen(cat_vars)", "_____no_output_____" ], [ "# make a list of the categorical variables that contain missing values\n\ncat_vars_with_na = [\n var for var in cat_vars\n if X_train[var].isnull().sum() > 0\n]\n\n# print percentage of missing values per variable\nX_train[cat_vars_with_na ].isnull().mean().sort_values(ascending=False)", "_____no_output_____" ], [ "# variables to impute with the string missing\nwith_string_missing = [\n var for var in cat_vars_with_na if X_train[var].isnull().mean() > 0.1]\n\n# variables to impute with the most frequent category\nwith_frequent_category = [\n var for var in cat_vars_with_na if X_train[var].isnull().mean() < 0.1]", "_____no_output_____" ], [ "with_string_missing", "_____no_output_____" ], [ "# replace missing values with new label: \"Missing\"\n\nX_train[with_string_missing] = X_train[with_string_missing].fillna('Missing')\nX_test[with_string_missing] = X_test[with_string_missing].fillna('Missing')", "_____no_output_____" ], [ "for var in with_frequent_category:\n \n # there can be more than 1 mode in a variable\n # we take the first one with [0] \n mode = X_train[var].mode()[0]\n \n print(var, mode)\n \n X_train[var].fillna(mode, inplace=True)\n X_test[var].fillna(mode, inplace=True)", "MasVnrType None\nBsmtQual TA\nBsmtCond TA\nBsmtExposure No\nBsmtFinType1 Unf\nBsmtFinType2 Unf\nElectrical SBrkr\nGarageType Attchd\nGarageFinish Unf\nGarageQual TA\nGarageCond TA\n" ], [ "# check that we have no missing information in the engineered variables\n\nX_train[cat_vars_with_na].isnull().sum()", "_____no_output_____" ], [ "# check that test set does not contain null values in the engineered variables\n\n[var for var in cat_vars_with_na if X_test[var].isnull().sum() > 0]", "_____no_output_____" ] ], [ [ "### Numerical variables\n\nTo engineer missing values in numerical variables, we will:\n\n- add a binary missing indicator variable\n- and then replace the missing values in the original variable with the mean", "_____no_output_____" ] ], [ [ "# now let's identify the numerical variables\n\nnum_vars = [\n var for var in X_train.columns if var not in cat_vars and var != 'SalePrice'\n]\n\n# number of numerical variables\nlen(num_vars)", "_____no_output_____" ], [ "# make a list with the numerical variables that contain missing values\nvars_with_na = [\n var for var in num_vars\n if X_train[var].isnull().sum() > 0\n]\n\n# print percentage of missing values per variable\nX_train[vars_with_na].isnull().mean()", "_____no_output_____" ], [ "# replace missing values as we described above\n\nfor var in vars_with_na:\n\n # calculate the mean using the train set\n mean_val = X_train[var].mean()\n \n print(var, mean_val)\n\n # add binary missing indicator (in train and test)\n X_train[var + '_na'] = np.where(X_train[var].isnull(), 1, 0)\n X_test[var + '_na'] = np.where(X_test[var].isnull(), 1, 0)\n\n # replace missing values by the mean\n # (in train and test)\n X_train[var].fillna(mean_val, inplace=True)\n X_test[var].fillna(mean_val, inplace=True)\n\n# check that we have no more missing values in the engineered variables\nX_train[vars_with_na].isnull().sum()", "LotFrontage 69.87974098057354\nMasVnrArea 103.7974006116208\nGarageYrBlt 1978.2959677419356\n" ], [ "# check that test set does not contain null values in the engineered variables\n\n[var for var in vars_with_na if X_test[var].isnull().sum() > 0]", "_____no_output_____" ], [ "# check the binary missing indicator variables\n\nX_train[['LotFrontage_na', 'MasVnrArea_na', 'GarageYrBlt_na']].head()", "_____no_output_____" ] ], [ [ "## Temporal variables\n\n### Capture elapsed time\n\nWe learned in the previous notebook, that there are 4 variables that refer to the years in which the house or the garage were built or remodeled. \n\nWe will capture the time elapsed between those variables and the year in which the house was sold:", "_____no_output_____" ] ], [ [ "def elapsed_years(df, var):\n # capture difference between the year variable\n # and the year in which the house was sold\n df[var] = df['YrSold'] - df[var]\n return df", "_____no_output_____" ], [ "for var in ['YearBuilt', 'YearRemodAdd', 'GarageYrBlt']:\n X_train = elapsed_years(X_train, var)\n X_test = elapsed_years(X_test, var)", "_____no_output_____" ], [ "# now we drop YrSold\nX_train.drop(['YrSold'], axis=1, inplace=True)\nX_test.drop(['YrSold'], axis=1, inplace=True)", "_____no_output_____" ] ], [ [ "## Numerical variable transformation\n\n### Logarithmic transformation\n\nIn the previous notebook, we observed that the numerical variables are not normally distributed.\n\nWe will transform with the logarightm the positive numerical variables in order to get a more Gaussian-like distribution.", "_____no_output_____" ] ], [ [ "for var in [\"LotFrontage\", \"1stFlrSF\", \"GrLivArea\"]:\n X_train[var] = np.log(X_train[var])\n X_test[var] = np.log(X_test[var])", "_____no_output_____" ], [ "# check that test set does not contain null values in the engineered variables\n[var for var in [\"LotFrontage\", \"1stFlrSF\", \"GrLivArea\"] if X_test[var].isnull().sum() > 0]", "_____no_output_____" ], [ "# same for train set\n[var for var in [\"LotFrontage\", \"1stFlrSF\", \"GrLivArea\"] if X_train[var].isnull().sum() > 0]", "_____no_output_____" ] ], [ [ "### Yeo-Johnson transformation\n\nWe will apply the Yeo-Johnson transformation to LotArea.", "_____no_output_____" ] ], [ [ "# the yeo-johnson transformation learns the best exponent to transform the variable\n# it needs to learn it from the train set: \nX_train['LotArea'], param = stats.yeojohnson(X_train['LotArea'])\n\n# and then apply the transformation to the test set with the same\n# parameter: see who this time we pass param as argument to the \n# yeo-johnson\nX_test['LotArea'] = stats.yeojohnson(X_test['LotArea'], lmbda=param)\n\nprint(param)", "-12.55283001172003\n" ], [ "# check absence of na in the train set\n[var for var in X_train.columns if X_train[var].isnull().sum() > 0]", "_____no_output_____" ], [ "# check absence of na in the test set\n[var for var in X_train.columns if X_test[var].isnull().sum() > 0]", "_____no_output_____" ] ], [ [ "### Binarize skewed variables\n\nThere were a few variables very skewed, we would transform those into binary variables.", "_____no_output_____" ] ], [ [ "skewed = [\n 'BsmtFinSF2', 'LowQualFinSF', 'EnclosedPorch',\n '3SsnPorch', 'ScreenPorch', 'MiscVal'\n]\n\nfor var in skewed:\n \n # map the variable values into 0 and 1\n X_train[var] = np.where(X_train[var]==0, 0, 1)\n X_test[var] = np.where(X_test[var]==0, 0, 1)", "_____no_output_____" ] ], [ [ "## Categorical variables\n\n### Apply mappings\n\nThese are variables which values have an assigned order, related to quality. For more information, check Kaggle website.", "_____no_output_____" ] ], [ [ "# re-map strings to numbers, which determine quality\n\nqual_mappings = {'Po': 1, 'Fa': 2, 'TA': 3, 'Gd': 4, 'Ex': 5, 'Missing': 0, 'NA': 0}\n\nqual_vars = ['ExterQual', 'ExterCond', 'BsmtQual', 'BsmtCond',\n 'HeatingQC', 'KitchenQual', 'FireplaceQu',\n 'GarageQual', 'GarageCond',\n ]\n\nfor var in qual_vars:\n X_train[var] = X_train[var].map(qual_mappings)\n X_test[var] = X_test[var].map(qual_mappings)", "_____no_output_____" ], [ "exposure_mappings = {'No': 1, 'Mn': 2, 'Av': 3, 'Gd': 4}\n\nvar = 'BsmtExposure'\n\nX_train[var] = X_train[var].map(exposure_mappings)\nX_test[var] = X_test[var].map(exposure_mappings)", "_____no_output_____" ], [ "finish_mappings = {'Missing': 0, 'NA': 0, 'Unf': 1, 'LwQ': 2, 'Rec': 3, 'BLQ': 4, 'ALQ': 5, 'GLQ': 6}\n\nfinish_vars = ['BsmtFinType1', 'BsmtFinType2']\n\nfor var in finish_vars:\n X_train[var] = X_train[var].map(finish_mappings)\n X_test[var] = X_test[var].map(finish_mappings)", "_____no_output_____" ], [ "garage_mappings = {'Missing': 0, 'NA': 0, 'Unf': 1, 'RFn': 2, 'Fin': 3}\n\nvar = 'GarageFinish'\n\nX_train[var] = X_train[var].map(garage_mappings)\nX_test[var] = X_test[var].map(garage_mappings)", "_____no_output_____" ], [ "fence_mappings = {'Missing': 0, 'NA': 0, 'MnWw': 1, 'GdWo': 2, 'MnPrv': 3, 'GdPrv': 4}\n\nvar = 'Fence'\n\nX_train[var] = X_train[var].map(fence_mappings)\nX_test[var] = X_test[var].map(fence_mappings)", "_____no_output_____" ], [ "# check absence of na in the train set\n[var for var in X_train.columns if X_train[var].isnull().sum() > 0]", "_____no_output_____" ] ], [ [ "### Removing Rare Labels\n\nFor the remaining categorical variables, we will group those categories that are present in less than 1% of the observations. That is, all values of categorical variables that are shared by less than 1% of houses, well be replaced by the string \"Rare\".\n\nTo learn more about how to handle categorical variables visit our course [Feature Engineering for Machine Learning](https://www.udemy.com/course/feature-engineering-for-machine-learning/?referralCode=A855148E05283015CF06) in Udemy.", "_____no_output_____" ] ], [ [ "# capture all quality variables\n\nqual_vars = qual_vars + finish_vars + ['BsmtExposure','GarageFinish','Fence']\n\n# capture the remaining categorical variables\n# (those that we did not re-map)\n\ncat_others = [\n var for var in cat_vars if var not in qual_vars\n]\n\nlen(cat_others)", "_____no_output_____" ], [ "def find_frequent_labels(df, var, rare_perc):\n \n # function finds the labels that are shared by more than\n # a certain % of the houses in the dataset\n\n df = df.copy()\n\n tmp = df.groupby(var)[var].count() / len(df)\n\n return tmp[tmp > rare_perc].index\n\n\nfor var in cat_others:\n \n # find the frequent categories\n frequent_ls = find_frequent_labels(X_train, var, 0.01)\n \n print(var, frequent_ls)\n print()\n \n # replace rare categories by the string \"Rare\"\n X_train[var] = np.where(X_train[var].isin(\n frequent_ls), X_train[var], 'Rare')\n \n X_test[var] = np.where(X_test[var].isin(\n frequent_ls), X_test[var], 'Rare')", "MSZoning Index(['FV', 'RH', 'RL', 'RM'], dtype='object', name='MSZoning')\n\nStreet Index(['Pave'], dtype='object', name='Street')\n\nAlley Index(['Grvl', 'Missing', 'Pave'], dtype='object', name='Alley')\n\nLotShape Index(['IR1', 'IR2', 'Reg'], dtype='object', name='LotShape')\n\nLandContour Index(['Bnk', 'HLS', 'Low', 'Lvl'], dtype='object', name='LandContour')\n\nUtilities Index(['AllPub'], dtype='object', name='Utilities')\n\nLotConfig Index(['Corner', 'CulDSac', 'FR2', 'Inside'], dtype='object', name='LotConfig')\n\nLandSlope Index(['Gtl', 'Mod'], dtype='object', name='LandSlope')\n\nNeighborhood Index(['Blmngtn', 'BrDale', 'BrkSide', 'ClearCr', 'CollgCr', 'Crawfor',\n 'Edwards', 'Gilbert', 'IDOTRR', 'MeadowV', 'Mitchel', 'NAmes', 'NWAmes',\n 'NoRidge', 'NridgHt', 'OldTown', 'SWISU', 'Sawyer', 'SawyerW',\n 'Somerst', 'StoneBr', 'Timber'],\n dtype='object', name='Neighborhood')\n\nCondition1 Index(['Artery', 'Feedr', 'Norm', 'PosN', 'RRAn'], dtype='object', name='Condition1')\n\nCondition2 Index(['Norm'], dtype='object', name='Condition2')\n\nBldgType Index(['1Fam', '2fmCon', 'Duplex', 'Twnhs', 'TwnhsE'], dtype='object', name='BldgType')\n\nHouseStyle Index(['1.5Fin', '1Story', '2Story', 'SFoyer', 'SLvl'], dtype='object', name='HouseStyle')\n\nRoofStyle Index(['Gable', 'Hip'], dtype='object', name='RoofStyle')\n\nRoofMatl Index(['CompShg'], dtype='object', name='RoofMatl')\n\nExterior1st Index(['AsbShng', 'BrkFace', 'CemntBd', 'HdBoard', 'MetalSd', 'Plywood',\n 'Stucco', 'VinylSd', 'Wd Sdng', 'WdShing'],\n dtype='object', name='Exterior1st')\n\nExterior2nd Index(['AsbShng', 'BrkFace', 'CmentBd', 'HdBoard', 'MetalSd', 'Plywood',\n 'Stucco', 'VinylSd', 'Wd Sdng', 'Wd Shng'],\n dtype='object', name='Exterior2nd')\n\nMasVnrType Index(['BrkFace', 'None', 'Stone'], dtype='object', name='MasVnrType')\n\nFoundation Index(['BrkTil', 'CBlock', 'PConc', 'Slab'], dtype='object', name='Foundation')\n\nHeating Index(['GasA', 'GasW'], dtype='object', name='Heating')\n\nCentralAir Index(['N', 'Y'], dtype='object', name='CentralAir')\n\nElectrical Index(['FuseA', 'FuseF', 'SBrkr'], dtype='object', name='Electrical')\n\nFunctional Index(['Min1', 'Min2', 'Mod', 'Typ'], dtype='object', name='Functional')\n\nGarageType Index(['Attchd', 'Basment', 'BuiltIn', 'Detchd'], dtype='object', name='GarageType')\n\nPavedDrive Index(['N', 'P', 'Y'], dtype='object', name='PavedDrive')\n\nPoolQC Index(['Missing'], dtype='object', name='PoolQC')\n\nMiscFeature Index(['Missing', 'Shed'], dtype='object', name='MiscFeature')\n\nSaleType Index(['COD', 'New', 'WD'], dtype='object', name='SaleType')\n\nSaleCondition Index(['Abnorml', 'Family', 'Normal', 'Partial'], dtype='object', name='SaleCondition')\n\nMSSubClass Int64Index([20, 30, 50, 60, 70, 75, 80, 85, 90, 120, 160, 190], dtype='int64', name='MSSubClass')\n\n" ] ], [ [ "### Encoding of categorical variables\n\nNext, we need to transform the strings of the categorical variables into numbers. \n\nWe will do it so that we capture the monotonic relationship between the label and the target.\n\nTo learn more about how to encode categorical variables visit our course [Feature Engineering for Machine Learning](https://www.udemy.com/course/feature-engineering-for-machine-learning/?referralCode=A855148E05283015CF06) in Udemy.", "_____no_output_____" ] ], [ [ "# this function will assign discrete values to the strings of the variables,\n# so that the smaller value corresponds to the category that shows the smaller\n# mean house sale price\n\ndef replace_categories(train, test, y_train, var, target):\n \n tmp = pd.concat([X_train, y_train], axis=1)\n \n # order the categories in a variable from that with the lowest\n # house sale price, to that with the highest\n ordered_labels = tmp.groupby([var])[target].mean().sort_values().index\n\n # create a dictionary of ordered categories to integer values\n ordinal_label = {k: i for i, k in enumerate(ordered_labels, 0)}\n \n print(var, ordinal_label)\n print()\n\n # use the dictionary to replace the categorical strings by integers\n train[var] = train[var].map(ordinal_label)\n test[var] = test[var].map(ordinal_label)", "_____no_output_____" ], [ "for var in cat_others:\n replace_categories(X_train, X_test, y_train, var, 'SalePrice')", "MSZoning {'Rare': 0, 'RM': 1, 'RH': 2, 'RL': 3, 'FV': 4}\n\nStreet {'Rare': 0, 'Pave': 1}\n\nAlley {'Grvl': 0, 'Pave': 1, 'Missing': 2}\n\nLotShape {'Reg': 0, 'IR1': 1, 'Rare': 2, 'IR2': 3}\n\nLandContour {'Bnk': 0, 'Lvl': 1, 'Low': 2, 'HLS': 3}\n\nUtilities {'Rare': 0, 'AllPub': 1}\n\nLotConfig {'Inside': 0, 'FR2': 1, 'Corner': 2, 'Rare': 3, 'CulDSac': 4}\n\nLandSlope {'Gtl': 0, 'Mod': 1, 'Rare': 2}\n\nNeighborhood {'IDOTRR': 0, 'MeadowV': 1, 'BrDale': 2, 'Edwards': 3, 'BrkSide': 4, 'OldTown': 5, 'Sawyer': 6, 'SWISU': 7, 'NAmes': 8, 'Mitchel': 9, 'SawyerW': 10, 'Rare': 11, 'NWAmes': 12, 'Gilbert': 13, 'Blmngtn': 14, 'CollgCr': 15, 'Crawfor': 16, 'ClearCr': 17, 'Somerst': 18, 'Timber': 19, 'StoneBr': 20, 'NridgHt': 21, 'NoRidge': 22}\n\nCondition1 {'Artery': 0, 'Feedr': 1, 'Norm': 2, 'RRAn': 3, 'Rare': 4, 'PosN': 5}\n\nCondition2 {'Rare': 0, 'Norm': 1}\n\nBldgType {'2fmCon': 0, 'Duplex': 1, 'Twnhs': 2, '1Fam': 3, 'TwnhsE': 4}\n\nHouseStyle {'SFoyer': 0, '1.5Fin': 1, 'Rare': 2, '1Story': 3, 'SLvl': 4, '2Story': 5}\n\nRoofStyle {'Gable': 0, 'Rare': 1, 'Hip': 2}\n\nRoofMatl {'CompShg': 0, 'Rare': 1}\n\nExterior1st {'AsbShng': 0, 'Wd Sdng': 1, 'WdShing': 2, 'MetalSd': 3, 'Stucco': 4, 'Rare': 5, 'HdBoard': 6, 'Plywood': 7, 'BrkFace': 8, 'CemntBd': 9, 'VinylSd': 10}\n\nExterior2nd {'AsbShng': 0, 'Wd Sdng': 1, 'MetalSd': 2, 'Wd Shng': 3, 'Stucco': 4, 'Rare': 5, 'HdBoard': 6, 'Plywood': 7, 'BrkFace': 8, 'CmentBd': 9, 'VinylSd': 10}\n\nMasVnrType {'Rare': 0, 'None': 1, 'BrkFace': 2, 'Stone': 3}\n\nFoundation {'Slab': 0, 'BrkTil': 1, 'CBlock': 2, 'Rare': 3, 'PConc': 4}\n\nHeating {'Rare': 0, 'GasW': 1, 'GasA': 2}\n\nCentralAir {'N': 0, 'Y': 1}\n\nElectrical {'Rare': 0, 'FuseF': 1, 'FuseA': 2, 'SBrkr': 3}\n\nFunctional {'Rare': 0, 'Min2': 1, 'Mod': 2, 'Min1': 3, 'Typ': 4}\n\nGarageType {'Rare': 0, 'Detchd': 1, 'Basment': 2, 'Attchd': 3, 'BuiltIn': 4}\n\nPavedDrive {'N': 0, 'P': 1, 'Y': 2}\n\nPoolQC {'Missing': 0, 'Rare': 1}\n\nMiscFeature {'Rare': 0, 'Shed': 1, 'Missing': 2}\n\nSaleType {'COD': 0, 'Rare': 1, 'WD': 2, 'New': 3}\n\nSaleCondition {'Rare': 0, 'Abnorml': 1, 'Family': 2, 'Normal': 3, 'Partial': 4}\n\nMSSubClass {30: 0, 'Rare': 1, 190: 2, 90: 3, 160: 4, 50: 5, 85: 6, 70: 7, 80: 8, 20: 9, 75: 10, 120: 11, 60: 12}\n\n" ], [ "# check absence of na in the train set\n[var for var in X_train.columns if X_train[var].isnull().sum() > 0]", "_____no_output_____" ], [ "# check absence of na in the test set\n[var for var in X_test.columns if X_test[var].isnull().sum() > 0]", "_____no_output_____" ], [ "# let me show you what I mean by monotonic relationship\n# between labels and target\n\ndef analyse_vars(train, y_train, var):\n \n # function plots median house sale price per encoded\n # category\n \n tmp = pd.concat([X_train, np.log(y_train)], axis=1)\n \n tmp.groupby(var)['SalePrice'].median().plot.bar()\n plt.title(var)\n plt.ylim(2.2, 2.6)\n plt.ylabel('SalePrice')\n plt.show()\n \nfor var in cat_others:\n analyse_vars(X_train, y_train, var)", "_____no_output_____" ] ], [ [ "The monotonic relationship is particularly clear for the variables MSZoning and Neighborhood. Note how, the higher the integer that now represents the category, the higher the mean house sale price.\n\n(remember that the target is log-transformed, that is why the differences seem so small).", "_____no_output_____" ], [ "## Feature Scaling\n\nFor use in linear models, features need to be either scaled. We will scale features to the minimum and maximum values:", "_____no_output_____" ] ], [ [ "# create scaler\nscaler = MinMaxScaler()\n\n# fit the scaler to the train set\nscaler.fit(X_train) \n\n# transform the train and test set\n\n# sklearn returns numpy arrays, so we wrap the\n# array with a pandas dataframe\n\nX_train = pd.DataFrame(\n scaler.transform(X_train),\n columns=X_train.columns\n)\n\nX_test = pd.DataFrame(\n scaler.transform(X_test),\n columns=X_train.columns\n)", "_____no_output_____" ], [ "X_train.head()", "_____no_output_____" ], [ "# let's now save the train and test sets for the next notebook!\n\nX_train.to_csv('xtrain.csv', index=False)\nX_test.to_csv('xtest.csv', index=False)\n\ny_train.to_csv('ytrain.csv', index=False)\ny_test.to_csv('ytest.csv', index=False)", "_____no_output_____" ], [ "# now let's save the scaler\n\njoblib.dump(scaler, 'minmax_scaler.joblib') ", "_____no_output_____" ] ], [ [ "That concludes the feature engineering section.\n\n# Additional Resources\n\n- [Feature Engineering for Machine Learning](https://www.udemy.com/course/feature-engineering-for-machine-learning/?referralCode=A855148E05283015CF06) - Online Course\n- [Packt Feature Engineering Cookbook](https://www.packtpub.com/data/python-feature-engineering-cookbook) - Book\n- [Feature Engineering for Machine Learning: A comprehensive Overview](https://trainindata.medium.com/feature-engineering-for-machine-learning-a-comprehensive-overview-a7ad04c896f8) - Article\n- [Practical Code Implementations of Feature Engineering for Machine Learning with Python](https://towardsdatascience.com/practical-code-implementations-of-feature-engineering-for-machine-learning-with-python-f13b953d4bcd) - Article", "_____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" ]

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

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]