markdown
stringlengths 0
1.02M
| code
stringlengths 0
832k
| output
stringlengths 0
1.02M
| license
stringlengths 3
36
| path
stringlengths 6
265
| repo_name
stringlengths 6
127
|
|---|---|---|---|---|---|
Uarray reduced Now let's use uarray's `optimize` decorator to create an updated function that specifes the dimensionality of the arrays to produced an optimized form: | # enable_logging()
optimized_some_fn = optimize(args[0].shape, args[1].shape)(some_fn) | _____no_output_____ | BSD-3-Clause | notebooks/NumPy Compat.ipynb | costrouc/uarray |
Now let's try our function out to see if it's faster: | # NBVAL_IGNORE_OUTPUT
%timeit optimized_some_fn(*args) | 5.47 µs ± 48.5 ns per loop (mean ± std. dev. of 7 runs, 100000 loops each)
| BSD-3-Clause | notebooks/NumPy Compat.ipynb | costrouc/uarray |
Yep about 10x as fast. Let's look at how this is done! First, we create an abstract representation of the array operations: | optimized_some_fn.__optimize_steps__['resulting_expr'] | _____no_output_____ | BSD-3-Clause | notebooks/NumPy Compat.ipynb | costrouc/uarray |
Then, we compile that to Python AST: | print(optimized_some_fn.__optimize_steps__['ast_as_source']) |
def fn(a, b):
i_5 = ()
i_6 = 10
i_1 = ((i_6,) + i_5)
i_0 = np.empty(i_1)
i_2 = 10
for i_3 in range(i_2):
i_4 = i_0[i_3]
i_9 = 5
i_10 = a
i_13 = i_10[i_9]
i_11 = i_3
i_12 = b
i_14 = i_12[i_11]
i_4 = (i_13 * i_14)
i_0[i_3] = i_4
return i_0
| BSD-3-Clause | notebooks/NumPy Compat.ipynb | costrouc/uarray |
Numba optimized To give this an extra speed boost, we can compile the returned expression with Numba: | numba_optimized = njit(optimized_some_fn)
# NBVAL_IGNORE_OUTPUT
# run once first to compile
numba_optimized(*args)
%timeit numba_optimized(*args) | 876 ns ± 16.7 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)
| BSD-3-Clause | notebooks/NumPy Compat.ipynb | costrouc/uarray |
Great, another speedup! Unkown dimensionality? What if we want to produce a version of the function that works on any dimensional input? Or if we just want to actually defer to NumPy's implementation and not replace `outer`? We simply omit the `with_dim` methods and we get back an abstract representation that is compiled without any knowledge of the dimensionality: | dims_not_known = optimize(some_fn)
dims_not_known.__optimize_steps__['resulting_expr']
print(dims_not_known.__optimize_steps__['ast_as_source']) |
def fn(a, b):
i_18 = 5
i_16 = a
i_17 = b
i_19 = np.multiply.outer(i_16, i_17)
i_15 = i_19[i_18]
return i_15
| BSD-3-Clause | notebooks/NumPy Compat.ipynb | costrouc/uarray |
TweepyPara acessar a API do twitter, deve-se acessar a área de desenvolvedor, criar uma aplicação e solicitar as credenciais de acesso. As células abaixo demonstram como fazer a conexão na API do twitter usando a biblioteca python Tweepy e retornar a lista de tweets na timeline do usuário.[Documentação do tweepy](https://tweepy.readthedocs.io/) | consumer_key = os.environ['CONSUMER_API_KEY']
consumer_secret = os.environ['CONSUMER_API_SECRET']
access_token = os.environ['ACCESS_TOKEN']
access_token_secret = os.environ['ACCESS_TOKEN_SECRET']
auth = tweepy.OAuthHandler(consumer_key, consumer_secret)
auth.set_access_token(access_token, access_token_secret)
api = tweepy.API(auth) | _____no_output_____ | Apache-2.0 | twitter-crawler.ipynb | phdabel/twitter-crawler |
Recuperar tweets da timeline | public_tweets = api.home_timeline() | _____no_output_____ | Apache-2.0 | twitter-crawler.ipynb | phdabel/twitter-crawler |
Criar um dicionário vazio para incluir os tweets.Passo necessário para criar um dataframe. | tweets = {'id':[],'author':[], 'screen_name':[], 'tweet':[],'created_at':[],'language':[],'n_retweets':[],'n_likes':[]}
ct = 0
num_results = 1000
result_count = 0
last_id = None
while result_count < num_results:
gremio_tweets = api.search(q='gremio',lang='pt',since_id=last_id)
for tweet in gremio_tweets:
print("id: "+tweet.id_str)
print("Screen name: "+tweet.author.screen_name)
print("Autor: "+tweet.author.name)
print("Tweet: "+tweet.text)
print("Data de criação: "+tweet.created_at.strftime("%d/%m/%Y, %H:%M:%S"))
print("Idioma: "+tweet.lang)
print("Retweets: "+str(tweet.retweet_count))
print("Curtidas: "+str(tweet.favorite_count))
tweets['id'].append(tweet.id)
tweets['screen_name'].append(tweet.author.screen_name)
tweets['author'].append(tweet.author.name)
tweets['tweet'].append(tweet.text)
tweets['created_at'].append(tweet.created_at)
tweets['language'].append(tweet.lang)
tweets['n_retweets'].append(tweet.retweet_count)
tweets['n_likes'].append(tweet.favorite_count)
print("==========================")
result_count += 1 | id: 1201655788508471298
Screen name: RauenPaian
Autor: IMORTAL 🖤💙🖤💙
Tweet: RT @SoccerGremio: A informação que nos chega, é que são apenas detalhes a serem finalizados entre Grêmio e Palmeiras por Raphael Veiga. Os…
Data de criação: 03/12/2019, 00:14:00
Idioma: pt
Retweets: 54
Curtidas: 0
==========================
id: 1201655782686760960
Screen name: regisschuch
Autor: Régis
Tweet: Esperamos pela parte do @Gremio Que não contrate esse Egídio. JOGADOR muito ruim.
Data de criação: 03/12/2019, 00:13:58
Idioma: pt
Retweets: 0
Curtidas: 0
==========================
id: 1201655773706833922
Screen name: magrin00
Autor: 🄼🄰🄶🅁🄸🄽ᶜʳᶠ ⚫🔴
Tweet: RT @TozzaFla: Fifa reconhece títulos mundiais de Flamengo, Grêmio, Santos e São Paulo | futebol internacional | Globoesporte https://t.co/A…
Data de criação: 03/12/2019, 00:13:56
Idioma: pt
Retweets: 232
Curtidas: 0
==========================
id: 1201655770200317958
Screen name: BaseFuracao
Autor: Base Furacão
Tweet: Em atuação coletiva, 3-0 vs Boca.
Em emoção e vibração, Remontada vs Grêmio. https://t.co/ebQwsoBUmo
Data de criação: 03/12/2019, 00:13:55
Idioma: pt
Retweets: 0
Curtidas: 0
==========================
id: 1201655768107356164
Screen name: Proerrd
Autor: Vivi pereiraⓟ
Tweet: @di_dinelli @isastrevisan comissão rainha grêmio nadinha
Data de criação: 03/12/2019, 00:13:55
Idioma: pt
Retweets: 0
Curtidas: 0
==========================
id: 1201655765859262465
Screen name: jean_wosch
Autor: Jean Wosch
Tweet: @omenguista @LibertadoresBR @AthleticoPR @Flamengo @Gremio @Palmeiras @SantosFC @SaoPauloFC Não reclama, se não pio… https://t.co/n60v1B2Mby
Data de criação: 03/12/2019, 00:13:54
Idioma: pt
Retweets: 0
Curtidas: 0
==========================
id: 1201655763921453056
Screen name: avalyzinho
Autor: nx.avaly 15/11
Tweet: RT @Jhorobert11: Obrigado senhor 🙏🏽🙏🏽
Feliz por mais uma vitória e pelos 2 gols ⚽️⚽️ Isso é Grêmio 🇪🇪🇪🇪
JR11 https://t.co/AEzxnn3GLt
Data de criação: 03/12/2019, 00:13:54
Idioma: pt
Retweets: 20
Curtidas: 0
==========================
id: 1201655762931666948
Screen name: DanielZ80970238
Autor: Daniel Zurita
Tweet: RT @TozzaFla: Fifa reconhece títulos mundiais de Flamengo, Grêmio, Santos e São Paulo | futebol internacional | Globoesporte https://t.co/A…
Data de criação: 03/12/2019, 00:13:54
Idioma: pt
Retweets: 232
Curtidas: 0
==========================
id: 1201655755826442241
Screen name: BianoRL1
Autor: @BianoRL
Tweet: RT @ejramorim: Gostaria de agradecer as atletas, comissão e direção por essa temporada, é um enorme prazer poder conviver e aprender diaria…
Data de criação: 03/12/2019, 00:13:52
Idioma: pt
Retweets: 3
Curtidas: 0
==========================
id: 1201655731537268736
Screen name: LuizCar74542646
Autor: Luiz Carlos
Tweet: Tu tens razão, qual era a meia cancha do grêmio em 2017, campeão da libertadores, quem entrou e nunca mais saiu Art… https://t.co/VNcZjOWna6
Data de criação: 03/12/2019, 00:13:46
Idioma: pt
Retweets: 0
Curtidas: 0
==========================
id: 1201655713065512960
Screen name: ceciliaraisa
Autor: O Chelsea kerrLUTE 😒💙😒💦
Tweet: @manamaiara @leilinha_1910 @tathiane_vidal pra não dar briga é melhor ele ir pro Grêmio
Data de criação: 03/12/2019, 00:13:42
Idioma: pt
Retweets: 0
Curtidas: 0
==========================
id: 1201655702562967552
Screen name: cxcarloos
Autor: carrlos
Tweet: RT @SoccerGremio: A informação que nos chega, é que são apenas detalhes a serem finalizados entre Grêmio e Palmeiras por Raphael Veiga. Os…
Data de criação: 03/12/2019, 00:13:39
Idioma: pt
Retweets: 54
Curtidas: 0
==========================
id: 1201655693784338433
Screen name: willianselong_
Autor: ₩illian
Tweet: @cesarspo @sandra_kunst @Gremio Vo passar por burro nada o comentário é meu e quem decide sou eu, evito ladainha de pessoas como fosse.
Data de criação: 03/12/2019, 00:13:37
Idioma: pt
Retweets: 0
Curtidas: 0
==========================
id: 1201655691825553411
Screen name: pmsilva37
Autor: Coalito 🐨
Tweet: @carlydamasceno2 Com empate também, caso o Grêmio vença o Cruzeiro.
Ficaria 3 pontos na frente e pelo número de vit… https://t.co/hA7TYBW0Jk
Data de criação: 03/12/2019, 00:13:37
Idioma: pt
Retweets: 0
Curtidas: 0
==========================
id: 1201655682900078592
Screen name: VNJS18
Autor: Vinicin
Tweet: RT @RDTRubroNegro: Flamengo pré Jorge Jesus não vencia:
- A Liberta há 38 anos.
- O Brasileiro há 10 anos.
- Na Arena da Baixada há 45 a…
Data de criação: 03/12/2019, 00:13:35
Idioma: pt
Retweets: 597
Curtidas: 0
==========================
id: 1201655788508471298
Screen name: RauenPaian
Autor: IMORTAL 🖤💙🖤💙
Tweet: RT @SoccerGremio: A informação que nos chega, é que são apenas detalhes a serem finalizados entre Grêmio e Palmeiras por Raphael Veiga. Os…
Data de criação: 03/12/2019, 00:14:00
Idioma: pt
Retweets: 54
Curtidas: 0
==========================
id: 1201655782686760960
Screen name: regisschuch
Autor: Régis
Tweet: Esperamos pela parte do @Gremio Que não contrate esse Egídio. JOGADOR muito ruim.
Data de criação: 03/12/2019, 00:13:58
Idioma: pt
Retweets: 0
Curtidas: 0
==========================
id: 1201655773706833922
Screen name: magrin00
Autor: 🄼🄰🄶🅁🄸🄽ᶜʳᶠ ⚫🔴
Tweet: RT @TozzaFla: Fifa reconhece títulos mundiais de Flamengo, Grêmio, Santos e São Paulo | futebol internacional | Globoesporte https://t.co/A…
Data de criação: 03/12/2019, 00:13:56
Idioma: pt
Retweets: 232
Curtidas: 0
==========================
id: 1201655770200317958
Screen name: BaseFuracao
Autor: Base Furacão
Tweet: Em atuação coletiva, 3-0 vs Boca.
Em emoção e vibração, Remontada vs Grêmio. https://t.co/ebQwsoBUmo
Data de criação: 03/12/2019, 00:13:55
Idioma: pt
Retweets: 0
Curtidas: 0
==========================
id: 1201655768107356164
Screen name: Proerrd
Autor: Vivi pereiraⓟ
Tweet: @di_dinelli @isastrevisan comissão rainha grêmio nadinha
Data de criação: 03/12/2019, 00:13:55
Idioma: pt
Retweets: 0
Curtidas: 0
==========================
id: 1201655765859262465
Screen name: jean_wosch
Autor: Jean Wosch
Tweet: @omenguista @LibertadoresBR @AthleticoPR @Flamengo @Gremio @Palmeiras @SantosFC @SaoPauloFC Não reclama, se não pio… https://t.co/n60v1B2Mby
Data de criação: 03/12/2019, 00:13:54
Idioma: pt
Retweets: 0
Curtidas: 0
==========================
id: 1201655763921453056
Screen name: avalyzinho
Autor: nx.avaly 15/11
Tweet: RT @Jhorobert11: Obrigado senhor 🙏🏽🙏🏽
Feliz por mais uma vitória e pelos 2 gols ⚽️⚽️ Isso é Grêmio 🇪🇪🇪🇪
JR11 https://t.co/AEzxnn3GLt
Data de criação: 03/12/2019, 00:13:54
Idioma: pt
Retweets: 20
Curtidas: 0
==========================
id: 1201655762931666948
Screen name: DanielZ80970238
Autor: Daniel Zurita
Tweet: RT @TozzaFla: Fifa reconhece títulos mundiais de Flamengo, Grêmio, Santos e São Paulo | futebol internacional | Globoesporte https://t.co/A…
Data de criação: 03/12/2019, 00:13:54
Idioma: pt
Retweets: 232
Curtidas: 0
==========================
id: 1201655755826442241
Screen name: BianoRL1
Autor: @BianoRL
Tweet: RT @ejramorim: Gostaria de agradecer as atletas, comissão e direção por essa temporada, é um enorme prazer poder conviver e aprender diaria…
Data de criação: 03/12/2019, 00:13:52
Idioma: pt
Retweets: 3
Curtidas: 0
==========================
id: 1201655731537268736
Screen name: LuizCar74542646
Autor: Luiz Carlos
Tweet: Tu tens razão, qual era a meia cancha do grêmio em 2017, campeão da libertadores, quem entrou e nunca mais saiu Art… https://t.co/VNcZjOWna6
Data de criação: 03/12/2019, 00:13:46
Idioma: pt
Retweets: 0
Curtidas: 0
==========================
id: 1201655713065512960
Screen name: ceciliaraisa
Autor: O Chelsea kerrLUTE 😒💙😒💦
Tweet: @manamaiara @leilinha_1910 @tathiane_vidal pra não dar briga é melhor ele ir pro Grêmio
Data de criação: 03/12/2019, 00:13:42
Idioma: pt
Retweets: 0
Curtidas: 0
==========================
id: 1201655702562967552
Screen name: cxcarloos
Autor: carrlos
Tweet: RT @SoccerGremio: A informação que nos chega, é que são apenas detalhes a serem finalizados entre Grêmio e Palmeiras por Raphael Veiga. Os…
Data de criação: 03/12/2019, 00:13:39
Idioma: pt
Retweets: 54
Curtidas: 0
==========================
id: 1201655693784338433
Screen name: willianselong_
Autor: ₩illian
Tweet: @cesarspo @sandra_kunst @Gremio Vo passar por burro nada o comentário é meu e quem decide sou eu, evito ladainha de pessoas como fosse.
Data de criação: 03/12/2019, 00:13:37
Idioma: pt
Retweets: 0
Curtidas: 0
==========================
id: 1201655691825553411
Screen name: pmsilva37
Autor: Coalito 🐨
Tweet: @carlydamasceno2 Com empate também, caso o Grêmio vença o Cruzeiro.
Ficaria 3 pontos na frente e pelo número de vit… https://t.co/hA7TYBW0Jk
Data de criação: 03/12/2019, 00:13:37
Idioma: pt
Retweets: 0
Curtidas: 0
==========================
id: 1201655682900078592
Screen name: VNJS18
Autor: Vinicin
Tweet: RT @RDTRubroNegro: Flamengo pré Jorge Jesus não vencia:
- A Liberta há 38 anos.
- O Brasileiro há 10 anos.
- Na Arena da Baixada há 45 a…
Data de criação: 03/12/2019, 00:13:35
Idioma: pt
Retweets: 597
Curtidas: 0
==========================
| Apache-2.0 | twitter-crawler.ipynb | phdabel/twitter-crawler |
Criação do dataframe | df = pd.DataFrame.from_dict(tweets)
df | _____no_output_____ | Apache-2.0 | twitter-crawler.ipynb | phdabel/twitter-crawler |
Salvar o dataframe em csv para uso posterior. | df.to_csv('dataframe.csv') | _____no_output_____ | Apache-2.0 | twitter-crawler.ipynb | phdabel/twitter-crawler |
Assignment 2Before working on this assignment please read these instructions fully. In the submission area, you will notice that you can click the link to **Preview the Grading** for each step of the assignment. This is the criteria that will be used for peer grading. Please familiarize yourself with the criteria before beginning the assignment.An NOAA dataset has been stored in the file `data/C2A2_data/BinnedCsvs_d100/4e86d2106d0566c6ad9843d882e72791333b08be3d647dcae4f4b110.csv`. The data for this assignment comes from a subset of The National Centers for Environmental Information (NCEI) [Daily Global Historical Climatology Network](https://www1.ncdc.noaa.gov/pub/data/ghcn/daily/readme.txt) (GHCN-Daily). The GHCN-Daily is comprised of daily climate records from thousands of land surface stations across the globe.Each row in the assignment datafile corresponds to a single observation.The following variables are provided to you:* **id** : station identification code* **date** : date in YYYY-MM-DD format (e.g. 2012-01-24 = January 24, 2012)* **element** : indicator of element type * TMAX : Maximum temperature (tenths of degrees C) * TMIN : Minimum temperature (tenths of degrees C)* **value** : data value for element (tenths of degrees C)For this assignment, you must:1. Read the documentation and familiarize yourself with the dataset, then write some python code which returns a line graph of the record high and record low temperatures by day of the year over the period 2005-2014. The area between the record high and record low temperatures for each day should be shaded.2. Overlay a scatter of the 2015 data for any points (highs and lows) for which the ten year record (2005-2014) record high or record low was broken in 2015.3. Watch out for leap days (i.e. February 29th), it is reasonable to remove these points from the dataset for the purpose of this visualization.4. Make the visual nice! Leverage principles from the first module in this course when developing your solution. Consider issues such as legends, labels, and chart junk.The data you have been given is near **Singapore, Central Singapore Community Development Council, Singapore**, and the stations the data comes from are shown on the map below. | import matplotlib.pyplot as plt
import mplleaflet
import pandas as pd
def leaflet_plot_stations(binsize, hashid):
df = pd.read_csv('data/C2A2_data/BinSize_d{}.csv'.format(binsize))
station_locations_by_hash = df[df['hash'] == hashid]
lons = station_locations_by_hash['LONGITUDE'].tolist()
lats = station_locations_by_hash['LATITUDE'].tolist()
plt.figure(figsize=(8,8))
plt.scatter(lons, lats, c='r', alpha=0.7, s=200)
return mplleaflet.display()
leaflet_plot_stations(100,'4e86d2106d0566c6ad9843d882e72791333b08be3d647dcae4f4b110')
# Import useful libraries
import matplotlib.pyplot as plt
import matplotlib.dates as dates
import matplotlib.ticker as ticker
import pandas as pd
import numpy as np
%matplotlib notebook
# Read the dataframe
df1 = pd.read_csv('data/C2A2_data/BinnedCsvs_d100/4e86d2106d0566c6ad9843d882e72791333b08be3d647dcae4f4b110.csv')
df1.head()
#How many records?
len(df1)
minimum = []
maximum = []
month = []
#remove February 29
df1 = df1[~(df1['Date'].str.endswith(r'02-29'))]
times1 = pd.DatetimeIndex(df1['Date'])
#after removing Feb 29, how many records remaining
len(df1)
#Data for 2005-2014
df = df1[times1.year != 2015]
times = pd.DatetimeIndex(df['Date'])
for j in df.groupby([times.month, times.day]):
minimum.append(min(j[1]['Data_Value']))
maximum.append(max(j[1]['Data_Value']))
#Data of 2015
df2015 = df1[times1.year == 2015]
times2015 = pd.DatetimeIndex(df2015['Date'])
minimum2015 = []
maximum2015 = []
for j in df2015.groupby([times2015.month, times2015.day]):
minimum2015.append(min(j[1]['Data_Value']))
maximum2015.append(max(j[1]['Data_Value']))
minaxis = []
maxaxis = []
minvals = []
maxvals = []
for i in range(len(minimum)):
if((minimum[i] - minimum2015[i]) > 0):
minaxis.append(i)
minvals.append(minimum2015[i])
if((maximum[i] - maximum2015[i]) < 0):
maxaxis.append(i)
maxvals.append(maximum2015[i])
plt.figure()
colors = ['skyblue', 'lightcoral']
plt.plot(minimum, c='skyblue', alpha = 0.5, label = 'Minimum Temperature (2005-14)')
plt.plot(maximum, c ='lightcoral', alpha = 0.5, label = 'Maximum Temperature (2005-14)')
plt.scatter(minaxis, minvals, s = 10, c = 'blue', label = 'Record Break Minimum (2015)')
plt.scatter(maxaxis, maxvals, s = 10, c = 'red', label = 'Record Break Maximum (2015)')
plt.gca().fill_between(range(len(minimum)),
minimum, maximum,
facecolor='lightgray',
alpha=0.2)
plt.ylim(0, 450)
plt.legend(loc = 8, frameon=False, title='Temperature', fontsize=8)
plt.xticks( np.linspace(15,15 + 30*11 , num = 12), (r'Jan', r'Feb', r'Mar', r'Apr', r'May', r'Jun', r'Jul', r'Aug', r'Sep', r'Oct', r'Nov', r'Dec') )
plt.xlabel('Months')
plt.ylabel('Temperature (tenths of degrees C)')
plt.title(r'Temperature Summary Plot of Singapore (2005-2015)')
plt.show()
plt.savefig('Temperature.png', transparent = True, bbox_inches = 'tight') | _____no_output_____ | MIT | Applied Data Science with Python Specialzation/Applied Plotting Charting and Data Representation in Python/Assignment2/Assignment2.ipynb | lynnxlmiao/Coursera |
Classical Machine Learning ApproachIn this notebook we will be learning to 1. Create a Naive TF - IDF based Bag of Words representation of text. 2. Use classical ML models to solve text classification. 3. Use a One Vs Rest strategy to solve multi-label text classification. **HOT TIP** : *Save them as pickle for easy rendering for experiments* This Notebook uses code from https://github.com/susanli2016/Machine-Learning-with-Python/blob/master/Multi%20label%20text%20classification.ipynb | # Installing packages.
!pip install contractions
!pip install textsearch
!pip install tqdm
# Importing packages.
import nltk
nltk.download('punkt')
nltk.download('stopwords')
%matplotlib inline
import re
import matplotlib
import numpy as np
import matplotlib.pyplot as plt
import pandas as pd
from sklearn.model_selection import train_test_split
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.naive_bayes import MultinomialNB
from sklearn.metrics import accuracy_score
from sklearn.multiclass import OneVsRestClassifier
from nltk.corpus import stopwords
stop_words = set(stopwords.words('english'))
from sklearn.svm import LinearSVC
from sklearn.linear_model import LogisticRegression
from sklearn.pipeline import Pipeline
import seaborn as sns
from sklearn.metrics import confusion_matrix, classification_report
import pickle
import ast
from sklearn.externals import joblib
from datetime import datetime
from sklearn.preprocessing import MultiLabelBinarizer
# Let's mount our G-Drive.
from google.colab import drive
drive.mount('/content/drive', force_remount=True)
# Data read and preparation.
# Mentioning where is our data located on G-Drive. Make sure to rectify your path
path = '/content/drive/My Drive/ICDMAI_Tutorial/notebook/'
data ='filtered_data/question_tag_text_mapping.pkl'
ml_model = path + 'ml_model/'
# Let us quickly load our question tag data
question_tag = pd.read_pickle(path+data)
question_tag.head(3) | _____no_output_____ | MIT | 2_classical_ml_approach.ipynb | funmilola09/Recurrent-Neural-Pipeline |
Creating one hot encoding from multilabelled tagged data | # In order to use one vs rest strategy we will need to one hot encoding each tag across all documents.
mlb = MultiLabelBinarizer()
question_tag['Tag_pop'] = question_tag['Tag']
question_tag = question_tag.join(pd.DataFrame(mlb.fit_transform(question_tag.pop('Tag_pop')),
columns=mlb.classes_,
index=question_tag.index))
question_tag.head(3)
# Creating a list of all existing 'Tags'
dummy = question_tag.drop(['Id', 'OwnerUserId', 'CreationDate', 'ClosedDate', 'Score', 'Title','Body','Tag'], axis=1)
categories = list(dummy.columns.values) | _____no_output_____ | MIT | 2_classical_ml_approach.ipynb | funmilola09/Recurrent-Neural-Pipeline |
Text preprocessing | # Let us createa a very basic text preprocessor which we will use for cleaning text.
def clean_text(text):
text = text.lower()
text = re.sub(r"what's", "what is ", text)
text = re.sub(r"\'s", " ", text)
text = re.sub(r"\'ve", " have ", text)
text = re.sub(r"can't", "can not ", text)
text = re.sub(r"n't", " not ", text)
text = re.sub(r"i'm", "i am ", text)
text = re.sub(r"\'re", " are ", text)
text = re.sub(r"\'d", " would ", text)
text = re.sub(r"\'ll", " will ", text)
text = re.sub(r"\'scuse", " excuse ", text)
text = re.sub('\W', ' ', text)
text = re.sub('\s+', ' ', text)
text = text.strip(' ')
return text
question_tag['Body'] = question_tag['Body'].map(lambda com : clean_text(com)) | _____no_output_____ | MIT | 2_classical_ml_approach.ipynb | funmilola09/Recurrent-Neural-Pipeline |
Creating a 70/30 Train-Test Split | train, test = train_test_split(question_tag, random_state=42, test_size=0.30, shuffle=True)
X_train = train.Body
X_test = test.Body
print("Train data shape : {}".format(X_train.shape))
print("Test data shape : {}".format(X_test.shape)) | Train data shape : (736394,)
Test data shape : (315598,)
| MIT | 2_classical_ml_approach.ipynb | funmilola09/Recurrent-Neural-Pipeline |
Creating Bag of Words representation using TF - IDF 1. Initializing the Vectorizer object 2. Create a corpus from training data. 3. Create a document term matrix | #Initializing the Vectorizer object
tfidf = TfidfVectorizer(stop_words=stop_words)
#Create a corpus from training data
#Create a document term matrix of training data based on the corpus.
X_train_dtm = tfidf.fit_transform(X_train)
#Create a document term matrix of test data based on the corpus.
#Note that the dimensions/columns of DTM of the test data will be based on the training data corpus only.
X_test_dtm = tfidf.transform(X_test) | _____no_output_____ | MIT | 2_classical_ml_approach.ipynb | funmilola09/Recurrent-Neural-Pipeline |
Pipelinescikit-learn provides a Pipeline utility to help automate machine learning workflows. Pipelines are very common in Machine Learning systems, since there is a lot of data to manipulate and many data transformations to apply. So we will utilize pipeline to train every classifier. One Vs Rest Multilabel strategyThe Multi-label algorithm accepts a binary mask over multiple labels. The result for each prediction will be an array of 0s and 1s marking which class labels apply to each row input sample.OneVsRest strategy can be used for multilabel learning, where a classifier is used to predict multiple labels for instance. **Naive Bayes**, **SVM**, **Logistic Regression** supports multi-class, but we are in a multi-label scenario, therefore, we wrap them in the OneVsRestClassifier. We create a Training Pipeline and a Scoring Pipeline | def tag_level_training_pipeline(X_train, train, X_test, test, classifier_pipeline, output_directory):
#1. Create a classifier for each Tag
for category in categories:
print('... Processing {}'.format(category))
# 1. train the model using X_dtm & y
classifier_pipeline.fit(X_train, train[category])
# 2. save the model to disk
filename = ml_model + output_directory +str(category)+ '_model.pkl'
joblib.dump(classifier_pipeline, filename, compress = 1)
# 3. compute the testing accuracy
prediction = classifier_pipeline.predict(X_test)
print('Test accuracy is {}'.format(accuracy_score(test[category], prediction)))
print(classification_report(test[category], prediction))
def tag_level_predict(X_train, train, X_test, test, model_directory):
prediction_df = pd.DataFrame(columns=['dummy1'])
#Score the document across classifier for each Tag
for category in categories:
# 1. load the model
filename = ml_model + model_directory +str(category)+ '_model.pkl'
classifier_pipeline = joblib.load(filename)
# 2. predict on the test data.
prediction = classifier_pipeline.predict(X_test)
prediction_df[str(category)] = prediction
# Remember We had encoded the labels. It time to bring them back to their original form.
for category in categories:
prediction_df.loc[prediction_df[str(category)] == 1, str(category)] = category
prediction_df['predicted_labels'] = prediction_df[[str(i) for i in categories]].values.tolist()
prediction_df['predicted_labels'] = prediction_df['predicted_labels'].apply(lambda x : list(set(x)))
# prediction_df['predicted_labels'] = prediction_df['predicted_labels'].apply(lambda x: x.remove(0) if (0 in x) else x )
# We create result having orignal labels and predicted labels for metrics Evaluation
final_pred_df = pd.concat([test[['Id','Tag']].reset_index(), prediction_df[['predicted_labels']].reset_index()], axis=1)
final_pred_df['original_labels'] = final_pred_df['Tag']
# prediction_df[['Id']] = test[['Id']]
final_pred_df_result = final_pred_df[['Id','original_labels','predicted_labels']]
return final_pred_df_result
# importing os module
import os
try:
os.rename('/content/drive/My Drive/ICDMAI_Tutorial/notebook/ml_model/SVM/_net_model.pkl', '/content/drive/My Drive/ICDMAI_Tutorial/notebook/ml_model/SVM/.net_model.pkl')
except :
print("Already in proper filename!")
## A Dummy example.
X_test = ["How to handle memory locking ?", "How to handle memory locking in java ?", "How to handle memory locking in java python ?","This post is not about java"]
X_test_dtm = tfidf.transform(X_test)
result = tag_level_predict(X_train_dtm, train, X_test_dtm, test.head(1), 'SVM/')
for i in range(result.shape[0]):
print("Input [",X_test[i],"] || Predicted classes: ",result.predicted_labels[i]) | /usr/local/lib/python3.6/dist-packages/sklearn/utils/deprecation.py:144: FutureWarning: The sklearn.svm.classes module is deprecated in version 0.22 and will be removed in version 0.24. The corresponding classes / functions should instead be imported from sklearn.svm. Anything that cannot be imported from sklearn.svm is now part of the private API.
warnings.warn(message, FutureWarning)
/usr/local/lib/python3.6/dist-packages/sklearn/base.py:318: UserWarning: Trying to unpickle estimator LinearSVC from version 0.21.3 when using version 0.22.1. This might lead to breaking code or invalid results. Use at your own risk.
UserWarning)
/usr/local/lib/python3.6/dist-packages/sklearn/utils/deprecation.py:144: FutureWarning: The sklearn.preprocessing.label module is deprecated in version 0.22 and will be removed in version 0.24. The corresponding classes / functions should instead be imported from sklearn.preprocessing. Anything that cannot be imported from sklearn.preprocessing is now part of the private API.
warnings.warn(message, FutureWarning)
/usr/local/lib/python3.6/dist-packages/sklearn/base.py:318: UserWarning: Trying to unpickle estimator LabelBinarizer from version 0.21.3 when using version 0.22.1. This might lead to breaking code or invalid results. Use at your own risk.
UserWarning)
/usr/local/lib/python3.6/dist-packages/sklearn/base.py:318: UserWarning: Trying to unpickle estimator OneVsRestClassifier from version 0.21.3 when using version 0.22.1. This might lead to breaking code or invalid results. Use at your own risk.
UserWarning)
/usr/local/lib/python3.6/dist-packages/sklearn/base.py:318: UserWarning: Trying to unpickle estimator Pipeline from version 0.21.3 when using version 0.22.1. This might lead to breaking code or invalid results. Use at your own risk.
UserWarning)
| MIT | 2_classical_ml_approach.ipynb | funmilola09/Recurrent-Neural-Pipeline |
Evaluating our results | # Here we define precision, recall, f1 measure at a single document level.
def document_evaluation_metrics(prd_grp,grp,metric="precision"):
pred_group = prd_grp
if 0 in pred_group: pred_group.remove(0)
group = grp
set_pred_group = set(pred_group)
set_group = set(group)
intrsct = set_group.intersection(set_pred_group)
accuracy = len(intrsct) / float(len(set_pred_group) if len(set_pred_group)>1 else 1)
recall = len(intrsct) / float(len(set_group) if len(set_group)>1 else 1)
if metric == "precision":
return accuracy
elif metric == "recall":
return recall
elif metric == "f1_measure":
if accuracy == 0 or recall == 0:
return 0
elif accuracy > 0 and recall >0 :
f1_measure = 2*accuracy*recall/(float(accuracy + recall))
return f1_measure
return -1
# Provide overall average stats and populate document level metrics.
def model_evaluation_stats(final_pred_df, model_name="default"):
final_pred_df['doc_precision'] = final_pred_df.apply(lambda x: document_evaluation_metrics(x.predicted_labels, x.original_labels, "precision"), axis=1)
final_pred_df['doc_recall'] = final_pred_df.apply(lambda x: document_evaluation_metrics(x.predicted_labels, x.original_labels, "recall"), axis=1)
final_pred_df['doc_f1_measure'] = final_pred_df.apply(lambda x: document_evaluation_metrics(x.predicted_labels, x.original_labels, "f1_measure"), axis=1)
print('Avearge precision across documents is {}'.format(final_pred_df['doc_precision'].mean()))
print('Avearge recall across documents is {}'.format(final_pred_df['doc_recall'].mean()))
print('Avearge f1 measure across documents is {}'.format(final_pred_df['doc_f1_measure'].mean()))
pickle.dump(final_pred_df, open(ml_model + model_name + ".pkl", 'wb'))
# final_pred_df.to_csv(ml_model + 'SVM_Tag_predictions.txt',sep='\t',index=False) | _____no_output_____ | MIT | 2_classical_ml_approach.ipynb | funmilola09/Recurrent-Neural-Pipeline |
Let us train, score and evaluate Naive Bayes | #Naive Bayes Classifier
NB_pipeline = Pipeline([
('clf', OneVsRestClassifier(MultinomialNB(
fit_prior=True, class_prior=None))),
])
tag_level_training_pipeline(X_train_dtm, train, X_test_dtm, test, NB_pipeline, 'NaiveBayes/')
result = tag_level_predict(X_train_dtm, train, X_test_dtm, test, 'NaiveBayes/')
model_evaluation_stats(result, "NaiveBayes") | _____no_output_____ | MIT | 2_classical_ml_approach.ipynb | funmilola09/Recurrent-Neural-Pipeline |
Let us train, score and evaluate Support Vector Machines | #SVM Classifier
SVC_pipeline = Pipeline([
('clf', OneVsRestClassifier(LinearSVC(), n_jobs=1)),
])
tag_level_training_pipeline(X_train_dtm, train, X_test_dtm, test, SVC_pipeline, 'SVM/')
result = tag_level_predict(X_train_dtm, train, X_test_dtm, test, 'SVM/')
model_evaluation_stats(result, "SVM") | ... Processing .net
Test accuracy is 0.9771893358006071
precision recall f1-score support
0 0.98 1.00 0.99 308362
1 0.51 0.09 0.15 7236
accuracy 0.98 315598
macro avg 0.75 0.54 0.57 315598
weighted avg 0.97 0.98 0.97 315598
... Processing agile
Test accuracy is 0.9999429654180318
precision recall f1-score support
0 1.00 1.00 1.00 315573
1 0.89 0.32 0.47 25
accuracy 1.00 315598
macro avg 0.94 0.66 0.74 315598
weighted avg 1.00 1.00 1.00 315598
... Processing ajax
Test accuracy is 0.9887356700612805
precision recall f1-score support
0 0.99 1.00 0.99 310952
1 0.70 0.41 0.52 4646
accuracy 0.99 315598
macro avg 0.84 0.71 0.76 315598
weighted avg 0.99 0.99 0.99 315598
... Processing amazon-web-services
Test accuracy is 0.9981780619649048
precision recall f1-score support
0 1.00 1.00 1.00 314643
1 0.79 0.55 0.65 955
accuracy 1.00 315598
macro avg 0.89 0.77 0.82 315598
weighted avg 1.00 1.00 1.00 315598
... Processing android
Test accuracy is 0.9829815144582665
precision recall f1-score support
0 0.99 1.00 0.99 288322
1 0.96 0.84 0.90 27276
accuracy 0.98 315598
macro avg 0.97 0.92 0.94 315598
weighted avg 0.98 0.98 0.98 315598
... Processing android-studio
Test accuracy is 0.9972401599503166
precision recall f1-score support
0 1.00 1.00 1.00 314589
1 0.67 0.27 0.38 1009
accuracy 1.00 315598
macro avg 0.84 0.63 0.69 315598
weighted avg 1.00 1.00 1.00 315598
... Processing angular2
Test accuracy is 0.9991064582158315
precision recall f1-score support
0 1.00 1.00 1.00 314898
1 0.94 0.64 0.76 700
accuracy 1.00 315598
macro avg 0.97 0.82 0.88 315598
weighted avg 1.00 1.00 1.00 315598
... Processing angularjs
Test accuracy is 0.9949904625504598
precision recall f1-score support
0 1.00 1.00 1.00 309420
1 0.93 0.80 0.86 6178
accuracy 0.99 315598
macro avg 0.96 0.90 0.93 315598
weighted avg 0.99 0.99 0.99 315598
... Processing apache
Test accuracy is 0.9952788040481879
precision recall f1-score support
0 1.00 1.00 1.00 313590
1 0.71 0.43 0.54 2008
accuracy 1.00 315598
macro avg 0.85 0.72 0.77 315598
weighted avg 0.99 1.00 0.99 315598
... Processing apache-spark
Test accuracy is 0.999404305477221
precision recall f1-score support
0 1.00 1.00 1.00 314985
1 0.93 0.75 0.83 613
accuracy 1.00 315598
macro avg 0.97 0.87 0.91 315598
weighted avg 1.00 1.00 1.00 315598
... Processing api
Test accuracy is 0.9951900835873485
precision recall f1-score support
0 1.00 1.00 1.00 314094
1 0.47 0.07 0.12 1504
accuracy 1.00 315598
macro avg 0.73 0.53 0.56 315598
weighted avg 0.99 1.00 0.99 315598
... Processing asp.net
Test accuracy is 0.9815905043758199
precision recall f1-score support
0 0.99 1.00 0.99 306664
1 0.79 0.48 0.60 8934
accuracy 0.98 315598
macro avg 0.89 0.74 0.79 315598
weighted avg 0.98 0.98 0.98 315598
... Processing asp.net-web-api
Test accuracy is 0.9985804726265692
precision recall f1-score support
0 1.00 1.00 1.00 314999
1 0.72 0.41 0.52 599
accuracy 1.00 315598
macro avg 0.86 0.71 0.76 315598
weighted avg 1.00 1.00 1.00 315598
... Processing azure
Test accuracy is 0.9987769250755708
precision recall f1-score support
0 1.00 1.00 1.00 314517
1 0.91 0.71 0.80 1081
accuracy 1.00 315598
macro avg 0.95 0.86 0.90 315598
weighted avg 1.00 1.00 1.00 315598
... Processing bash
Test accuracy is 0.995047497132428
precision recall f1-score support
0 1.00 1.00 1.00 313311
1 0.76 0.46 0.58 2287
accuracy 1.00 315598
macro avg 0.88 0.73 0.79 315598
weighted avg 0.99 1.00 0.99 315598
... Processing c
Test accuracy is 0.9872432651664459
precision recall f1-score support
0 0.99 1.00 0.99 308691
1 0.81 0.55 0.65 6907
accuracy 0.99 315598
macro avg 0.90 0.77 0.82 315598
weighted avg 0.99 0.99 0.99 315598
... Processing c#
Test accuracy is 0.9414159785549971
precision recall f1-score support
0 0.95 0.98 0.97 285147
1 0.77 0.56 0.65 30451
accuracy 0.94 315598
macro avg 0.86 0.77 0.81 315598
weighted avg 0.94 0.94 0.94 315598
... Processing c++
Test accuracy is 0.9785581657678439
precision recall f1-score support
0 0.98 0.99 0.99 301367
1 0.85 0.63 0.73 14231
accuracy 0.98 315598
macro avg 0.92 0.81 0.86 315598
weighted avg 0.98 0.98 0.98 315598
... Processing cloud
Test accuracy is 0.9995722406352385
precision recall f1-score support
0 1.00 1.00 1.00 315459
1 0.75 0.04 0.08 139
accuracy 1.00 315598
macro avg 0.87 0.52 0.54 315598
weighted avg 1.00 1.00 1.00 315598
... Processing codeigniter
Test accuracy is 0.9979055634066122
precision recall f1-score support
0 1.00 1.00 1.00 314150
1 0.90 0.61 0.73 1448
accuracy 1.00 315598
macro avg 0.95 0.80 0.86 315598
weighted avg 1.00 1.00 1.00 315598
... Processing css
Test accuracy is 0.9785454914162954
precision recall f1-score support
0 0.99 0.99 0.99 302936
1 0.78 0.64 0.71 12662
accuracy 0.98 315598
macro avg 0.88 0.82 0.85 315598
weighted avg 0.98 0.98 0.98 315598
... Processing devops
Test accuracy is 0.9999524711816932
| MIT | 2_classical_ml_approach.ipynb | funmilola09/Recurrent-Neural-Pipeline |
Let us train, score and evaluate Logistic Regression | #Logistic Regression Classifier
LogReg_pipeline = Pipeline([
('clf', OneVsRestClassifier(LogisticRegression(solver='sag'), n_jobs=1)),
])
tag_level_training_pipeline(X_train_dtm, train, X_test_dtm, test, LogReg_pipeline, 'LogisticRegression/')
result = tag_level_predict(X_train_dtm, train, X_test_dtm, test, 'LogisticRegression/')
model_evaluation_stats(result, "LogisticRegression") | _____no_output_____ | MIT | 2_classical_ml_approach.ipynb | funmilola09/Recurrent-Neural-Pipeline |
Wikipedia Thanks-Receiver Study Randomization [J. Nathan Matias](https://twitter.com/natematias)October 29, 2019This code takes as input data described in the [randomization data format](https://docs.google.com/document/d/1plhoDbQryYQ32vZMXu8YmlLSp30QTdup43k6uTePOT4/edit?usp=drive_web&ouid=117701977297551627494) and produces randomizations for the Thanks Recipient study.Notes:* We use the 99% confidence interval cutoffs from our first sample rather than relative to each subsequent sample * Polish Experienced: 235.380736142341 * Polish Newcomer: 72.2118047599678 * Arabic Newcomer: 54.7365066602131 * German Newcomer: 63.3678642498622* We will be drawing only 300 Polish accounts | options("scipen"=9, "digits"=4)
library(ggplot2)
library(rlang)
library(tidyverse)
library(viridis)
library(blockTools)
library(blockrand)
library(gmodels) # contains CrossTable
library(DeclareDesign)
library(DescTools) # contains Freq
library(uuid)
options(repr.plot.width=7, repr.plot.height=3.5)
sessionInfo() | _____no_output_____ | MIT | randomization/thanks-recipient-study-2019/generate-wikipedia-thanks-recipient-randomizations-final-10.29.2019.R.ipynb | mitmedialab/CivilServant-Wikipedia-Analysis |
Load Input Dataframe | filename <- "all-thankees-historical-20191029.csv"
data.path <- "/home/civilservant/Tresors/CivilServant/projects/wikipedia-integration/gratitude-study/Data Drills/thankee"
recipient.df <- read.csv(file.path(data.path, "historical_output", filename)) | _____no_output_____ | MIT | randomization/thanks-recipient-study-2019/generate-wikipedia-thanks-recipient-randomizations-final-10.29.2019.R.ipynb | mitmedialab/CivilServant-Wikipedia-Analysis |
Load Participants in the Thanker Study | thanker.df <- read.csv(file.path(data.path, "..", "thanker_hardlaunch", "randomization_output",
"all-thanker-randomization-final-20190729.csv"))
usernames.to.exclude <- thanker.df$user_name | _____no_output_____ | MIT | randomization/thanks-recipient-study-2019/generate-wikipedia-thanks-recipient-randomizations-final-10.29.2019.R.ipynb | mitmedialab/CivilServant-Wikipedia-Analysis |
Load Liaison Usernames | liaison.df <- read.csv(file.path(data.path, "..", "thanker_hardlaunch", "randomization_output",
"liason-thanker-randomization-datadrill-20190718.csv"))
usernames.to.exclude <- append(as.character(usernames.to.exclude), as.character(liaison.df$user_name))
print(paste(length(usernames.to.exclude), "usernames to exclude")) | [1] "462 usernames to exclude"
| MIT | randomization/thanks-recipient-study-2019/generate-wikipedia-thanks-recipient-randomizations-final-10.29.2019.R.ipynb | mitmedialab/CivilServant-Wikipedia-Analysis |
Adjust Column Names to Match Thankee Randomization Specification | recipient.df$prev_experience <- factor(as.integer(gsub("bin_", "", recipient.df$prev_experience)))
recipient.df$anonymized_id <- sapply( seq_along(1:nrow(recipient.df)), UUIDgenerate )
recipient.df$newcomer <- recipient.df$prev_experience == 0
recipient.df <- subset(recipient.df, lang!="en")
#recipient.df <- subset(recipient.df, user_editcount_quality >=4 )
hist(recipient.df$user_editcount_quality) | _____no_output_____ | MIT | randomization/thanks-recipient-study-2019/generate-wikipedia-thanks-recipient-randomizations-final-10.29.2019.R.ipynb | mitmedialab/CivilServant-Wikipedia-Analysis |
Confirm the number of participants | print("Newcomer Participants to Randomize")
summary(subset(recipient.df, newcomer == 1)$lang)
## Polish Experienced Accounts
print("Experienced Participants to Randomize")
summary(subset(recipient.df, newcomer == 0)$lang) | [1] "Experienced Participants to Randomize"
| MIT | randomization/thanks-recipient-study-2019/generate-wikipedia-thanks-recipient-randomizations-final-10.29.2019.R.ipynb | mitmedialab/CivilServant-Wikipedia-Analysis |
Omit Participants Omit Participants in the Thanker Study | print(paste(nrow(recipient.df), "participants before removing thankers"))
recipient.df <- subset(recipient.df, (user_name %in% usernames.to.exclude)!=TRUE)
print(paste(nrow(recipient.df), "participants after removing thankers")) | [1] "3262 participants before removing thankers"
[1] "3262 participants after removing thankers"
| MIT | randomization/thanks-recipient-study-2019/generate-wikipedia-thanks-recipient-randomizations-final-10.29.2019.R.ipynb | mitmedialab/CivilServant-Wikipedia-Analysis |
Subset values outside the 99% confidence intervalsWe are using upper confidence intervals from the first randomization, found at [generate-wikipedia-thanks-recipient-randomizations-final-07.28.3019](generate-wikipedia-thanks-recipient-randomizations-final-07.28.3019.R.ipynb) * Polish Experienced: 235.380736142341 * Polish Newcomer: 72.2118047599678 * Arabic Newcomer: 54.7365066602131 * German Newcomer: 63.3678642498622 | upper.conf.ints <- data.frame(lang=c("pl", "pl", "de", "ar"),
newcomer=c(0,1,1,1),
conf.int = c(
235.380736142341,
72.2118047599678,
54.7365066602131,
63.3678642498622
))
upper.conf.ints
#subset(upper.conf.ints, lang=="pl" & newcomer ==1)$conf.int
## CREATE A PLACEHOLDER WITH ZERO ROWS
## BEFORE ITERATING
recipient.trimmed.df <- recipient.df[0,]
for(l in c("ar", "de", "fa", "pl")){
print(paste("Language: ", l))
for(n in c(0,1)){
print(paste(" newcomer:", n == 1))
lang.df <- subset(recipient.df, lang==l & newcomer == n)
print(paste( " ", nrow(lang.df), "rows from original dataset"))
prev.conf.int <- subset(upper.conf.ints, lang==l & newcomer ==n)$conf.int
print( " 99% confidence intervals:")
print(paste(" upper: ", prev.conf.int ,sep=""))
print(paste(" Removing",
nrow(subset(lang.df,
labor_hours_84_days_pre_sample > prev.conf.int)), "outliers",
"observations because labor_hours_84_days_pre_sample is an outlier."))
lang.subset.df <- subset(lang.df, labor_hours_84_days_pre_sample <= prev.conf.int)
print(paste( " ", nrow(lang.subset.df), "rows in trimmed dataset"))
recipient.trimmed.df <- rbind(recipient.trimmed.df, lang.subset.df)
}
}
recipient.df.penultimate <- recipient.trimmed.df | [1] "Language: ar"
[1] " newcomer: FALSE"
[1] " 0 rows from original dataset"
[1] " 99% confidence intervals:"
[1] " upper: "
[1] " Removing 0 outliers observations because labor_hours_84_days_pre_sample is an outlier."
[1] " 0 rows in trimmed dataset"
[1] " newcomer: TRUE"
[1] " 743 rows from original dataset"
[1] " 99% confidence intervals:"
[1] " upper: 63.3678642498622"
[1] " Removing 8 outliers observations because labor_hours_84_days_pre_sample is an outlier."
[1] " 735 rows in trimmed dataset"
[1] "Language: de"
[1] " newcomer: FALSE"
[1] " 0 rows from original dataset"
[1] " 99% confidence intervals:"
[1] " upper: "
[1] " Removing 0 outliers observations because labor_hours_84_days_pre_sample is an outlier."
[1] " 0 rows in trimmed dataset"
[1] " newcomer: TRUE"
[1] " 1565 rows from original dataset"
[1] " 99% confidence intervals:"
[1] " upper: 54.7365066602131"
[1] " Removing 37 outliers observations because labor_hours_84_days_pre_sample is an outlier."
[1] " 1528 rows in trimmed dataset"
[1] "Language: fa"
[1] " newcomer: FALSE"
[1] " 0 rows from original dataset"
[1] " 99% confidence intervals:"
[1] " upper: "
[1] " Removing 0 outliers observations because labor_hours_84_days_pre_sample is an outlier."
[1] " 0 rows in trimmed dataset"
[1] " newcomer: TRUE"
[1] " 0 rows from original dataset"
[1] " 99% confidence intervals:"
[1] " upper: "
[1] " Removing 0 outliers observations because labor_hours_84_days_pre_sample is an outlier."
[1] " 0 rows in trimmed dataset"
[1] "Language: pl"
[1] " newcomer: FALSE"
[1] " 512 rows from original dataset"
[1] " 99% confidence intervals:"
[1] " upper: 235.380736142341"
[1] " Removing 1 outliers observations because labor_hours_84_days_pre_sample is an outlier."
[1] " 511 rows in trimmed dataset"
[1] " newcomer: TRUE"
[1] " 442 rows from original dataset"
[1] " 99% confidence intervals:"
[1] " upper: 72.2118047599678"
[1] " Removing 7 outliers observations because labor_hours_84_days_pre_sample is an outlier."
[1] " 435 rows in trimmed dataset"
| MIT | randomization/thanks-recipient-study-2019/generate-wikipedia-thanks-recipient-randomizations-final-10.29.2019.R.ipynb | mitmedialab/CivilServant-Wikipedia-Analysis |
Review and Generate Variables | print(aggregate(recipient.df.penultimate[c("labor_hours_84_days_pre_sample")],
FUN=mean, by = list(recipient.df.penultimate$prev_experience)))
print(CrossTable(recipient.df.penultimate$has_email, recipient.df.penultimate$newcomer,
prop.r = FALSE, prop.c=TRUE, prop.t = FALSE, prop.chisq = FALSE))
## Update the has_email field
## recipient.df.penultimate$has_email <- recipient.df.penultimate$has_email == "True"
## PREVIOUS EXPERIENCE
print("prev_experience")
print(summary(factor(recipient.df.penultimate$prev_experience)))
cat("\n")
## SHOW LABOR HOURS BY EXPERIENCE GROUP:
print("Aggregate labor_hours_84_days_pre_sample")
print(aggregate(recipient.df.penultimate[c("labor_hours_84_days_pre_sample")],
FUN=mean, by = list(recipient.df.penultimate$prev_experience)))
cat("\n")
print("NEWCOMERS AND EMAILS")
print("--------------------")
print(CrossTable(recipient.df.penultimate$has_email, recipient.df.penultimate$newcomer,
prop.r = FALSE, prop.c=TRUE, prop.t = FALSE, prop.chisq = FALSE))
# VARIABLE: num_prev_thanks_pre_treatment
print("num_prev_thanks_pre_sample")
print(summary(recipient.df.penultimate$num_prev_thanks_pre_sample))
cat("\n")
## SHOW PREVIOUS THANKS BY EXPERIENCE GROUP:
print("num_prev_thanks_pre_sample by prev_experience")
print(aggregate(recipient.df.penultimate[c("num_prev_thanks_pre_sample")],
FUN=mean, by = list(recipient.df.penultimate$prev_experience)))
cat("\n") | [1] "prev_experience"
0 90 180 365 730 1460 2920
2698 63 52 69 81 102 144
[1] "Aggregate labor_hours_84_days_pre_sample"
Group.1 labor_hours_84_days_pre_sample
1 0 5.878
2 90 4.519
3 180 8.063
4 365 5.797
5 730 5.354
6 1460 5.866
7 2920 8.791
[1] "NEWCOMERS AND EMAILS"
[1] "--------------------"
Cell Contents
|-------------------------|
| N |
| N / Col Total |
|-------------------------|
Total Observations in Table: 3209
| recipient.df.penultimate$newcomer
recipient.df.penultimate$has_email | FALSE | TRUE | Row Total |
-----------------------------------|-----------|-----------|-----------|
False | 20 | 18 | 38 |
| 0.039 | 0.007 | |
-----------------------------------|-----------|-----------|-----------|
True | 491 | 2680 | 3171 |
| 0.961 | 0.993 | |
-----------------------------------|-----------|-----------|-----------|
Column Total | 511 | 2698 | 3209 |
| 0.159 | 0.841 | |
-----------------------------------|-----------|-----------|-----------|
$t
y
x FALSE TRUE
False 20 18
True 491 2680
$prop.row
y
x FALSE TRUE
False 0.5263 0.4737
True 0.1548 0.8452
$prop.col
y
x FALSE TRUE
False 0.039139 0.006672
True 0.960861 0.993328
$prop.tbl
y
x FALSE TRUE
False 0.006232 0.005609
True 0.153007 0.835151
[1] "num_prev_thanks_pre_sample"
Min. 1st Qu. Median Mean 3rd Qu. Max.
0.00 0.00 0.00 0.34 0.00 112.00
[1] "num_prev_thanks_pre_sample by prev_experience"
Group.1 num_prev_thanks_pre_sample
1 0 0.1542
2 90 0.1111
3 180 0.1731
4 365 0.2899
5 730 0.9630
6 1460 0.6176
7 2920 3.4097
| MIT | randomization/thanks-recipient-study-2019/generate-wikipedia-thanks-recipient-randomizations-final-10.29.2019.R.ipynb | mitmedialab/CivilServant-Wikipedia-Analysis |
Subset Sample to Planned sample sizesSample sizes are reported in the experiment [Decisions Document](https://docs.google.com/document/d/1HryhsmWI6WthXQC7zv9Hz1a9DhpZ3FxVRLjTONuMg4I/edit)* Arabic newcomers (1750 goal) (hoping for as many as possible in first sample) * hoping for 1350 in the first sample and 400 later* German newcomers (3000 goal) (hoping for as many as possible in first sample) * hoping for 1600 in first sample and 1400 later* Persian Experienced (2400 goal)* Polish: * Newcomers: (800 goal) * Experienced: (2400 goal) | ## Seed generated by Brooklyn Integers
# https://www.brooklynintegers.com/int/1495265601/
set.seed(1495265601)
print("Newcomers")
summary(subset(recipient.df.penultimate, newcomer==1)$lang)
print("Experienced")
summary(subset(recipient.df.penultimate, newcomer==0)$lang)
## CREATE THE FINAL PARTICIPANT SAMPLE BEFORE RANDOMIZATION
recipient.df.final <- recipient.df.penultimate | _____no_output_____ | MIT | randomization/thanks-recipient-study-2019/generate-wikipedia-thanks-recipient-randomizations-final-10.29.2019.R.ipynb | mitmedialab/CivilServant-Wikipedia-Analysis |
Generate Randomization Blocks | recipient.df.final$lang_prev_experience <- factor(paste(recipient.df.final$lang, recipient.df.final$prev_experience))
colnames(recipient.df.final)
## BLOCKING VARIABLES
bv = c("labor_hours_84_days_pre_sample", "num_prev_thanks_pre_sample")
block.size = 2
## TODO: CHECK TO SEE IF I CAN DO BALANCED RANDOMIZATION
## WITHIN BLOCKS LARGER THAN 2
blockobj = block(data=recipient.df.final,
n.tr = block.size,
groups = "lang_prev_experience",
id.vars="anonymized_id",
block.vars = bv,
distance ="mahalanobis"
)
## CHECK DISTANCES
#print(blockobj)
recipient.df.final$randomization_block_id <- createBlockIDs(blockobj,
data=recipient.df.final,
id.var = "anonymized_id")
recipient.df.final$randomization_block_size = block.size | _____no_output_____ | MIT | randomization/thanks-recipient-study-2019/generate-wikipedia-thanks-recipient-randomizations-final-10.29.2019.R.ipynb | mitmedialab/CivilServant-Wikipedia-Analysis |
Identify Incomplete Blocks and Remove Participants in Incomplete Blocks From the Experiment | block.sizes <- aggregate(recipient.df.final$randomization_block_id, FUN=length, by=list(recipient.df.final$randomization_block_id))
incomplete.blocks <- subset(block.sizes, x == 1)$Group.1
incomplete.blocks
nrow(subset(recipient.df.final, randomization_block_id %in% incomplete.blocks))
removed.observations <- subset(recipient.df.final, (
randomization_block_id %in% incomplete.blocks)==TRUE)
recipient.df.final <-
subset(recipient.df.final, (
randomization_block_id %in% incomplete.blocks)!=TRUE)
print(paste("Removed", nrow(removed.observations), "units placed in incomplete blocks.")) | [1] "Removed 5 units placed in incomplete blocks."
| MIT | randomization/thanks-recipient-study-2019/generate-wikipedia-thanks-recipient-randomizations-final-10.29.2019.R.ipynb | mitmedialab/CivilServant-Wikipedia-Analysis |
Generate Randomizations | assignments <- block_ra(blocks=recipient.df.final$randomization_block_id,
num_arms = 2, conditions = c(0,1))
recipient.df.final$randomization_arm <- assignments | _____no_output_____ | MIT | randomization/thanks-recipient-study-2019/generate-wikipedia-thanks-recipient-randomizations-final-10.29.2019.R.ipynb | mitmedialab/CivilServant-Wikipedia-Analysis |
Check Balance | print("Aggregating labor hours by treatment")
print(aggregate(recipient.df.final[c("labor_hours_84_days_pre_sample")],
FUN=mean, by = list(recipient.df.final$randomization_arm)))
print("CrossTable of lang by treatment")
CrossTable(recipient.df.final$lang, recipient.df.final$randomization_arm,
prop.r = TRUE, prop.c=FALSE, prop.t = FALSE, prop.chisq = FALSE)
print("CrossTable of lang_prev_experience by treatment")
CrossTable(recipient.df.final$lang_prev_experience, recipient.df.final$randomization_arm,
prop.r = TRUE, prop.c=FALSE, prop.t = FALSE, prop.chisq = FALSE)
| [1] "Aggregating labor hours by treatment"
Group.1 labor_hours_84_days_pre_sample
1 0 6.008
2 1 5.944
[1] "CrossTable of lang by treatment"
Cell Contents
|-------------------------|
| N |
| N / Row Total |
|-------------------------|
Total Observations in Table: 3204
| recipient.df.final$randomization_arm
recipient.df.final$lang | 0 | 1 | Row Total |
------------------------|-----------|-----------|-----------|
ar | 367 | 367 | 734 |
| 0.500 | 0.500 | 0.229 |
------------------------|-----------|-----------|-----------|
de | 764 | 764 | 1528 |
| 0.500 | 0.500 | 0.477 |
------------------------|-----------|-----------|-----------|
pl | 471 | 471 | 942 |
| 0.500 | 0.500 | 0.294 |
------------------------|-----------|-----------|-----------|
Column Total | 1602 | 1602 | 3204 |
------------------------|-----------|-----------|-----------|
[1] "CrossTable of lang_prev_experience by treatment"
Cell Contents
|-------------------------|
| N |
| N / Row Total |
|-------------------------|
Total Observations in Table: 3204
| recipient.df.final$randomization_arm
recipient.df.final$lang_prev_experience | 0 | 1 | Row Total |
----------------------------------------|-----------|-----------|-----------|
ar 0 | 367 | 367 | 734 |
| 0.500 | 0.500 | 0.229 |
----------------------------------------|-----------|-----------|-----------|
de 0 | 764 | 764 | 1528 |
| 0.500 | 0.500 | 0.477 |
----------------------------------------|-----------|-----------|-----------|
pl 0 | 217 | 217 | 434 |
| 0.500 | 0.500 | 0.135 |
----------------------------------------|-----------|-----------|-----------|
pl 1460 | 51 | 51 | 102 |
| 0.500 | 0.500 | 0.032 |
----------------------------------------|-----------|-----------|-----------|
pl 180 | 26 | 26 | 52 |
| 0.500 | 0.500 | 0.016 |
----------------------------------------|-----------|-----------|-----------|
pl 2920 | 72 | 72 | 144 |
| 0.500 | 0.500 | 0.045 |
----------------------------------------|-----------|-----------|-----------|
pl 365 | 34 | 34 | 68 |
| 0.500 | 0.500 | 0.021 |
----------------------------------------|-----------|-----------|-----------|
pl 730 | 40 | 40 | 80 |
| 0.500 | 0.500 | 0.025 |
----------------------------------------|-----------|-----------|-----------|
pl 90 | 31 | 31 | 62 |
| 0.500 | 0.500 | 0.019 |
----------------------------------------|-----------|-----------|-----------|
Column Total | 1602 | 1602 | 3204 |
----------------------------------------|-----------|-----------|-----------|
| MIT | randomization/thanks-recipient-study-2019/generate-wikipedia-thanks-recipient-randomizations-final-10.29.2019.R.ipynb | mitmedialab/CivilServant-Wikipedia-Analysis |
Subset Polish Experienced AccountsWithin Polish, identify 300 accounts (150 blocks) to include and drop all of the others.Note: since the previous randomization included a larger number of more experienced accounts, we're prioritizing accounts from experience groups 90, 180, 365, 730, and 1460 (all except 2920). | ## SHOW PREVIOUS THANKS BY EXPERIENCE GROUP:
recipient.df.final$count.var <- 1
print("Number of Accounts for each experience level among Polish Participants")
print(aggregate(subset(recipient.df.final, lang="pl")[c("count.var")],
FUN=sum, by = list(subset(recipient.df.final, lang="pl")$prev_experience)))
cat("\n")
print(paste("Total number of rows: ", nrow(recipient.df.final), sep=""))
recipient.df.final.a <- subset(recipient.df.final, !(lang=="pl" & prev_experience==2920))
recipient.df.final.a$initial.block.id <- recipient.df.final.a$randomization_block_id
print(paste("Total number of rows once we subset Polish:", nrow(recipient.df.final.a))) | [1] "Total number of rows: 3204"
[1] "Total number of rows once we subset Polish: 3060"
| MIT | randomization/thanks-recipient-study-2019/generate-wikipedia-thanks-recipient-randomizations-final-10.29.2019.R.ipynb | mitmedialab/CivilServant-Wikipedia-Analysis |
Offset block IDs to be uniqueObserve the block IDs from the previous randomizations and ensure that these ones are unique and larger. | ## LOAD PREVIOUS RANDOMIZATIONS
prev_randomization_filename <- "thanks-recipient-randomizations-20190729.csv"
prev.randomization.df <- read.csv(file.path(data.path, "randomization_output", prev_randomization_filename))
print(paste("Max Block ID: ", max(prev.randomization.df$randomization_block_id)))
prev.max.block.id <- max(prev.randomization.df$randomization_block_id)
prev.max.block.id <- 4221
recipient.df.final.a$randomization_block_id <- recipient.df.final.a$initial.block.id + prev.max.block.id
summary(recipient.df.final.a$randomization_block_id) | _____no_output_____ | MIT | randomization/thanks-recipient-study-2019/generate-wikipedia-thanks-recipient-randomizations-final-10.29.2019.R.ipynb | mitmedialab/CivilServant-Wikipedia-Analysis |
Sort by block ID | recipient.df.final.a <- recipient.df.final.a[order(recipient.df.final.a$randomization_block_id),]
print("Newcomers")
summary(subset(recipient.df.final.a, newcomer==1)$lang)
print("Experienced")
summary(subset(recipient.df.final.a, newcomer==0)$lang) | [1] "Newcomers"
| MIT | randomization/thanks-recipient-study-2019/generate-wikipedia-thanks-recipient-randomizations-final-10.29.2019.R.ipynb | mitmedialab/CivilServant-Wikipedia-Analysis |
Output and Archive Randomizations | randomization.filename <- paste("thanks-recipient-randomizations-", format(Sys.Date(), format="%Y%m%d"), ".csv", sep="")
write.csv(recipient.df.final.a, file = file.path(data.path, "randomization_output", randomization.filename))
colnames(recipient.df.final.a) | _____no_output_____ | MIT | randomization/thanks-recipient-study-2019/generate-wikipedia-thanks-recipient-randomizations-final-10.29.2019.R.ipynb | mitmedialab/CivilServant-Wikipedia-Analysis |
Get marriages | filepath = '/Volumes/backup_128G/z_repository/TBIO_data/RequestsFromTana/20190515'
filename = 'marriages.tsv'
read_filename = '{0}/{1}'.format(filepath, filename)
marriageDf = pd.read_csv(read_filename, delimiter='\t')
marriageDf.fillna('', inplace=True)
marriageDf.shape, marriageDf.head() | _____no_output_____ | MIT | tbio/marriages.ipynb | VincentCheng34/StudyOnPython |
To unique spouses | startId = 10000
personIds = {}
marriageDic = {}
for idx in range(0, len(marriageDf)):
row = marriageDf.loc[idx]
man = str(row['?manVal'])
wife = str(row['?wifeVal'])
woman = str(row['?womanVal'])
husband = str(row['?husbandVal'])
if man != '' and man not in personIds:
personIds[man] = startId
startId += 1
if wife != '' and wife not in personIds:
personIds[wife] = startId
startId += 1
if woman != '' and woman not in personIds:
personIds[woman] = startId
startId += 1
if husband != '' and husband not in personIds:
personIds[husband] = startId
startId += 1
if man not in marriageDic:
marriageDic[man] = [wife]
elif wife not in marriageDic[man]:
marriageDic[man].append(wife)
# else:
# print("man WRONG:", man, wife)
if wife not in marriageDic:
marriageDic[wife] = [man]
elif man not in marriageDic[wife]:
marriageDic[wife].append(man)
# else:
# print("wife WRONG:", wife, man)
if woman not in marriageDic:
marriageDic[woman] = [husband]
elif husband not in marriageDic[woman]:
marriageDic[woman].append(husband)
# else:
# print("woman WRONG:", woman, husband)
if husband not in marriageDic:
marriageDic[husband] = [woman]
elif woman not in marriageDic[husband]:
marriageDic[husband].append(woman)
# else:
# print("husband WRONG:", husband, woman)
# marriageDic
len(personIds)
personDf = pd.DataFrame(personIds, index=['ID']).T
personDf.head()
write_nodes_to = '{0}/{1}'.format(filepath, 'nodes_person_20190516_v2.xlsx')
personDf.to_excel(write_nodes_to) | _____no_output_____ | MIT | tbio/marriages.ipynb | VincentCheng34/StudyOnPython |
Read person-family map table | familymembers = 'Familymembers.xlsx'
read_familymembers = '{0}/{1}'.format(filepath, familymembers)
fmDf = pd.read_excel(read_familymembers)
fmDf.shape, fmDf.head()
startId = 20000
familyIds = {}
fmDic = {}
for idx in range(0, len(fmDf)):
row = fmDf.loc[idx]
person = str(row['personStr'])
family = str(row['familyStr'])
if family not in familyIds:
familyIds[family] = startId
startId += 1
if person not in fmDic:
fmDic[person] = family
elif family != fmDic[person]:
print("Dup:", person, family, fmDic[person])
# fmDic
len(familyIds)
familyDf = pd.DataFrame(familyIds, index=['ID']).T
familyDf.head()
write_nodes_to = '{0}/{1}'.format(filepath, 'nodes_family_20190516_v2.xlsx')
familyDf.to_excel(write_nodes_to) | _____no_output_____ | MIT | tbio/marriages.ipynb | VincentCheng34/StudyOnPython |
results `Source (Family)` | `Target(Family)`| `Type(Undirected)` | `Person/Source` | `Person/target` | def getFamilyName(INperson):
if INperson not in fmDic:
# print(INperson, " Not found!")
return ''
return fmDic[INperson]
def getPersonId(INperson):
if INperson not in personIds:
return 0
return personIds[INperson]
def getFamilyId(INfamily):
if INfamily not in familyIds:
return 0
return familyIds[INfamily]
resList = []
for sPerson in marriageDic:
spouses = marriageDic[sPerson]
for tPerson in spouses:
sFamily = getFamilyName(sPerson)
tFamily = getFamilyName(tPerson)
if sFamily == '' or tFamily == '':
# print(fPerson, fFamily, sPerson, sFamily)
continue
sPersonId = getPersonId(sPerson)
tPersonId = getPersonId(tPerson)
sFamilyId = getFamilyId(sFamily)
tFamilyId = getFamilyId(tFamily)
resList.append([sFamilyId, tFamilyId, 'Undirected', sPersonId, tPersonId])
print(len(resList))
resDf = pd.DataFrame(resList, columns=['SourceFamily', 'TargetFamily', 'Type',
'SourcePerson', 'TargetPerson'])
resDf.drop_duplicates(keep='first', inplace=True)
resDf.sort_values(by=['SourceFamily', 'TargetFamily', 'SourcePerson', 'TargetPerson'], inplace=True)
resDf.head()
print(len(resDf))
write_file_to = '{0}/{1}'.format(filepath, 'marriages_20190516_v2.xlsx')
resDf.to_excel(write_file_to, index=False) | _____no_output_____ | MIT | tbio/marriages.ipynb | VincentCheng34/StudyOnPython |
LeNet Lab SolutionSource: Yan LeCun Load DataLoad the MNIST data, which comes pre-loaded with TensorFlow.You do not need to modify this section. | from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets("MNIST_data/", reshape=False)
X_train, y_train = mnist.train.images, mnist.train.labels
X_validation, y_validation = mnist.validation.images, mnist.validation.labels
X_test, y_test = mnist.test.images, mnist.test.labels
assert(len(X_train) == len(y_train))
assert(len(X_validation) == len(y_validation))
assert(len(X_test) == len(y_test))
print()
print("Image Shape: {}".format(X_train[0].shape))
print()
print("Training Set: {} samples".format(len(X_train)))
print("Validation Set: {} samples".format(len(X_validation)))
print("Test Set: {} samples".format(len(X_test))) | G:\ProgramData\Anaconda3\envs\tensorflow\lib\site-packages\h5py\__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.
from ._conv import register_converters as _register_converters
| MIT | LeNet-Lab-Solution.ipynb | mdeopujari/CarND-Traffic-Sign-Classifier-Project |
The MNIST data that TensorFlow pre-loads comes as 28x28x1 images.However, the LeNet architecture only accepts 32x32xC images, where C is the number of color channels.In order to reformat the MNIST data into a shape that LeNet will accept, we pad the data with two rows of zeros on the top and bottom, and two columns of zeros on the left and right (28+2+2 = 32).You do not need to modify this section. | import numpy as np
# Pad images with 0s
X_train = np.pad(X_train, ((0,0),(2,2),(2,2),(0,0)), 'constant')
X_validation = np.pad(X_validation, ((0,0),(2,2),(2,2),(0,0)), 'constant')
X_test = np.pad(X_test, ((0,0),(2,2),(2,2),(0,0)), 'constant')
print("Updated Image Shape: {}".format(X_train[0].shape)) | Updated Image Shape: (36, 36, 1)
| MIT | LeNet-Lab-Solution.ipynb | mdeopujari/CarND-Traffic-Sign-Classifier-Project |
Visualize DataView a sample from the dataset.You do not need to modify this section. | import random
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline
index = random.randint(0, len(X_train))
image = X_train[index].squeeze()
plt.figure(figsize=(1,1))
plt.imshow(image, cmap="gray")
print(y_train[index]) | (32, 32)
4
| MIT | LeNet-Lab-Solution.ipynb | mdeopujari/CarND-Traffic-Sign-Classifier-Project |
Preprocess DataShuffle the training data.You do not need to modify this section. | from sklearn.utils import shuffle
X_train, y_train = shuffle(X_train, y_train) | _____no_output_____ | MIT | LeNet-Lab-Solution.ipynb | mdeopujari/CarND-Traffic-Sign-Classifier-Project |
Setup TensorFlowThe `EPOCH` and `BATCH_SIZE` values affect the training speed and model accuracy.You do not need to modify this section. | import tensorflow as tf
EPOCHS = 10
BATCH_SIZE = 128 | _____no_output_____ | MIT | LeNet-Lab-Solution.ipynb | mdeopujari/CarND-Traffic-Sign-Classifier-Project |
SOLUTION: Implement LeNet-5Implement the [LeNet-5](http://yann.lecun.com/exdb/lenet/) neural network architecture.This is the only cell you need to edit. InputThe LeNet architecture accepts a 32x32xC image as input, where C is the number of color channels. Since MNIST images are grayscale, C is 1 in this case. Architecture**Layer 1: Convolutional.** The output shape should be 28x28x6.**Activation.** Your choice of activation function.**Pooling.** The output shape should be 14x14x6.**Layer 2: Convolutional.** The output shape should be 10x10x16.**Activation.** Your choice of activation function.**Pooling.** The output shape should be 5x5x16.**Flatten.** Flatten the output shape of the final pooling layer such that it's 1D instead of 3D. The easiest way to do is by using `tf.contrib.layers.flatten`, which is already imported for you.**Layer 3: Fully Connected.** This should have 120 outputs.**Activation.** Your choice of activation function.**Layer 4: Fully Connected.** This should have 84 outputs.**Activation.** Your choice of activation function.**Layer 5: Fully Connected (Logits).** This should have 10 outputs. OutputReturn the result of the 2nd fully connected layer. | from tensorflow.contrib.layers import flatten
def LeNet(x):
# Arguments used for tf.truncated_normal, randomly defines variables for the weights and biases for each layer
mu = 0
sigma = 0.1
# SOLUTION: Layer 1: Convolutional. Input = 32x32x1. Output = 28x28x6.
conv1_W = tf.Variable(tf.truncated_normal(shape=(5, 5, 1, 6), mean = mu, stddev = sigma))
conv1_b = tf.Variable(tf.zeros(6))
conv1 = tf.nn.conv2d(x, conv1_W, strides=[1, 1, 1, 1], padding='VALID') + conv1_b
# SOLUTION: Activation.
conv1 = tf.nn.relu(conv1)
# SOLUTION: Pooling. Input = 28x28x6. Output = 14x14x6.
conv1 = tf.nn.max_pool(conv1, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='VALID')
# SOLUTION: Layer 2: Convolutional. Output = 10x10x16.
conv2_W = tf.Variable(tf.truncated_normal(shape=(5, 5, 6, 16), mean = mu, stddev = sigma))
conv2_b = tf.Variable(tf.zeros(16))
conv2 = tf.nn.conv2d(conv1, conv2_W, strides=[1, 1, 1, 1], padding='VALID') + conv2_b
# SOLUTION: Activation.
conv2 = tf.nn.relu(conv2)
# SOLUTION: Pooling. Input = 10x10x16. Output = 5x5x16.
conv2 = tf.nn.max_pool(conv2, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='VALID')
# SOLUTION: Flatten. Input = 5x5x16. Output = 400.
fc0 = flatten(conv2)
# SOLUTION: Layer 3: Fully Connected. Input = 400. Output = 120.
fc1_W = tf.Variable(tf.truncated_normal(shape=(400, 120), mean = mu, stddev = sigma))
fc1_b = tf.Variable(tf.zeros(120))
fc1 = tf.matmul(fc0, fc1_W) + fc1_b
# SOLUTION: Activation.
fc1 = tf.nn.relu(fc1)
# SOLUTION: Layer 4: Fully Connected. Input = 120. Output = 84.
fc2_W = tf.Variable(tf.truncated_normal(shape=(120, 84), mean = mu, stddev = sigma))
fc2_b = tf.Variable(tf.zeros(84))
fc2 = tf.matmul(fc1, fc2_W) + fc2_b
# SOLUTION: Activation.
fc2 = tf.nn.relu(fc2)
# SOLUTION: Layer 5: Fully Connected. Input = 84. Output = 10.
fc3_W = tf.Variable(tf.truncated_normal(shape=(84, 10), mean = mu, stddev = sigma))
fc3_b = tf.Variable(tf.zeros(10))
logits = tf.matmul(fc2, fc3_W) + fc3_b
return logits | _____no_output_____ | MIT | LeNet-Lab-Solution.ipynb | mdeopujari/CarND-Traffic-Sign-Classifier-Project |
Features and LabelsTrain LeNet to classify [MNIST](http://yann.lecun.com/exdb/mnist/) data.`x` is a placeholder for a batch of input images.`y` is a placeholder for a batch of output labels.You do not need to modify this section. | x = tf.placeholder(tf.float32, (None, 32, 32, 1))
y = tf.placeholder(tf.int32, (None))
one_hot_y = tf.one_hot(y, 10) | _____no_output_____ | MIT | LeNet-Lab-Solution.ipynb | mdeopujari/CarND-Traffic-Sign-Classifier-Project |
Training PipelineCreate a training pipeline that uses the model to classify MNIST data.You do not need to modify this section. | rate = 0.001
logits = LeNet(x)
cross_entropy = tf.nn.softmax_cross_entropy_with_logits(labels=one_hot_y, logits=logits)
loss_operation = tf.reduce_mean(cross_entropy)
optimizer = tf.train.AdamOptimizer(learning_rate = rate)
training_operation = optimizer.minimize(loss_operation) | _____no_output_____ | MIT | LeNet-Lab-Solution.ipynb | mdeopujari/CarND-Traffic-Sign-Classifier-Project |
Model EvaluationEvaluate how well the loss and accuracy of the model for a given dataset.You do not need to modify this section. | correct_prediction = tf.equal(tf.argmax(logits, 1), tf.argmax(one_hot_y, 1))
accuracy_operation = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
saver = tf.train.Saver()
def evaluate(X_data, y_data):
num_examples = len(X_data)
total_accuracy = 0
sess = tf.get_default_session()
for offset in range(0, num_examples, BATCH_SIZE):
batch_x, batch_y = X_data[offset:offset+BATCH_SIZE], y_data[offset:offset+BATCH_SIZE]
accuracy = sess.run(accuracy_operation, feed_dict={x: batch_x, y: batch_y})
total_accuracy += (accuracy * len(batch_x))
return total_accuracy / num_examples | _____no_output_____ | MIT | LeNet-Lab-Solution.ipynb | mdeopujari/CarND-Traffic-Sign-Classifier-Project |
Train the ModelRun the training data through the training pipeline to train the model.Before each epoch, shuffle the training set.After each epoch, measure the loss and accuracy of the validation set.Save the model after training.You do not need to modify this section. | with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
num_examples = len(X_train)
print("Training...")
print()
for i in range(EPOCHS):
X_train, y_train = shuffle(X_train, y_train)
for offset in range(0, num_examples, BATCH_SIZE):
end = offset + BATCH_SIZE
batch_x, batch_y = X_train[offset:end], y_train[offset:end]
sess.run(training_operation, feed_dict={x: batch_x, y: batch_y})
validation_accuracy = evaluate(X_validation, y_validation)
print("EPOCH {} ...".format(i+1))
print("Validation Accuracy = {:.3f}".format(validation_accuracy))
print()
saver.save(sess, './lenet')
print("Model saved") | _____no_output_____ | MIT | LeNet-Lab-Solution.ipynb | mdeopujari/CarND-Traffic-Sign-Classifier-Project |
Evaluate the ModelOnce you are completely satisfied with your model, evaluate the performance of the model on the test set.Be sure to only do this once!If you were to measure the performance of your trained model on the test set, then improve your model, and then measure the performance of your model on the test set again, that would invalidate your test results. You wouldn't get a true measure of how well your model would perform against real data.You do not need to modify this section. | with tf.Session() as sess:
saver.restore(sess, tf.train.latest_checkpoint('.'))
test_accuracy = evaluate(X_test, y_test)
print("Test Accuracy = {:.3f}".format(test_accuracy)) | _____no_output_____ | MIT | LeNet-Lab-Solution.ipynb | mdeopujari/CarND-Traffic-Sign-Classifier-Project |
Get data for only hepatocytes and rerun harmony+umap | # Reprocess t cell data
for ds in DS_LIST:
print(ds)
ind_select = dic_data_raw[ds].obs['cell_ontology_class']=='hepatocyte'
adata = dic_data_raw[ds][ind_select,:].copy()
sc.pp.filter_cells(adata, min_genes=250)
sc.pp.filter_genes(adata, min_cells=50)
adata.obs['batch_harmony'] = adata.obs['mouse.id']
adata.obs['batch_harmony'] = adata.obs['batch_harmony'].astype('category')
sc.pp.highly_variable_genes(adata, subset = False, min_disp=.5,
min_mean=.0125, max_mean=10, n_bins=20, n_top_genes=None)
sc.pp.scale(adata, max_value=10, zero_center=False)
sc.pp.pca(adata, n_comps=50, use_highly_variable=True, svd_solver='arpack')
sc.external.pp.harmony_integrate(adata, key='batch_harmony', max_iter_harmony=20)
sc.pp.neighbors(adata, n_neighbors=50, n_pcs=20, use_rep="X_pca_harmony")
# sc.pp.neighbors(adata, n_neighbors=50, n_pcs=20, use_rep="X_pca")
sc.tl.leiden(adata, resolution=0.7)
sc.tl.umap(adata)
sc.pl.umap(adata, color='cell_ontology_class')
sc.pl.umap(adata, color='leiden')
sc.pl.umap(adata, color=['age', 'sex', 'mouse.id', 'n_genes', 'subtissue'])
adata.write('/n/holystore01/LABS/price_lab/Users/mjzhang/scTRS_data/single_cell_data/tms_proc/'
'hep.%s.h5ad'%ds)
adata = read_h5ad('/n/holystore01/LABS/price_lab/Users/mjzhang/scTRS_data/single_cell_data/tms_proc/hep.droplet.h5ad')
# dic_cluster = {'0':'4', '1':'3','2':'1','3':'5','4':'0','5':'2'}
# adata.obs['leiden_old'] = adata.obs['leiden'].values
# adata.obs['leiden'] = [dic_cluster[x] for x in adata.obs['leiden_old']]
adata.obs = adata.obs.join(dic_score['droplet'])
sc.pl.umap(adata, color='leiden')
sc.pl.umap(adata, color='UKB_460K.biochemistry_LDLdirect.norm_score')
# sc.pl.umap(adata, color=['Pecam1', 'Nrp1', 'Kdr', 'Oit3'])
# sc.pl.umap(adata, color=['Clec4f', 'Cd68', 'Irf7'])
# sc.pl.umap(adata, color=['Alb', 'Ttr', 'Apoa1', 'Serpina1c'])
# adata.write('/n/holystore01/LABS/price_lab/Users/mjzhang/scTRS_data/single_cell_data/tms_proc/hep.facs_annot.h5ad')
temp_df = dic_data_raw['facs'][dic_data_raw['facs'].obs['cell_ontology_class']=='hepatocyte']\
.obs.groupby(['subtissue', 'mouse.id']).agg({'cell':len})
temp_df.loc[~temp_df['cell'].isna()] | _____no_output_____ | MIT | experiments/job.case_hepatocyte/s1_reprocess_tms_droplet_hep.ipynb | martinjzhang/scDRS |
Dot Placeholder Pull() and dot behaves in the same way | library(tidyverse)
library(dslabs)
data(murders)
murders <- murders %>% mutate(murder_rate = (total/population) * 100000)
summarize(murders, mean = mean(murder_rate))
us_murder_rate <- murders %>% summarize(rate = sum(total)/ sum(population) *100000)
r <- us_murder_rate %>% .$rate
class(r)
r
class(us_murder_rate$rate)
us_murder_rate <- murders %>%
summarize(rate = sum(total) / sum(population) * 100000) %>%
.$rate
us_murder_rate
heights %>% group_by(sex) %>% summarize(mean = mean(height), sd = sd(height))
murders %>% mutate(murder_rate = total/population * 100000) %>% group_by(region) %>% summarize(mean = mean(murder_rate), sd = sd(murder_rate))
murders %>% mutate(murder_rate = total/population * 100000) %>% arrange(desc(murder_rate)) %>% top_n(10, murder_rate)
murders %>% arrange(population) %>% head()
murders %>% arrange(desc(population)) %>% head()
murders %>% arrange(region, population) %>% head()
murders %>% top_n(10, murder_rate) | _____no_output_____ | MIT | HAR_DM/Practice Work/Summarizing with dplyr.ipynb | ashudva/HAR |
Gaussian Mixture Models (GMM)KDE centers each bin (or kernel rather) at each point. In a [**mixture model**](https://en.wikipedia.org/wiki/Mixture_model) we don't use a kernel for each data point, but rather we fit for the *locations of the kernels*--in addition to the width. So a mixture model is sort of a hybrid between an $N$-D histogram and KDE. Using lots of kernels (maybe even more than the BIC score suggests) may make sense if you just want to provide an accurate description of the data (as in density estimation). Using fewer kernels makes mixture models more like clustering, where the suggestion is still to use many kernels in order to divide the sample into real clusters and "background". Gaussians are the most commonly used components for mixture models. So, the pdf is modeled by a sum of Gaussians:$$p(x) = \sum_{k=1}^N \alpha_k \mathscr{N}(x|\mu_k,\Sigma_k),$$where $\alpha_k$ is the "mixing coefficient" with $0\le \alpha_k \le 1$ and $\sum_{k=1}^N \alpha_k = 1$.We can solve for the parameters using maximum likelihood analyis as we have discussed previously.However, this can be complicated in multiple dimensions, requiring the use of [**Expectation Maximization (EM)**](https://en.wikipedia.org/wiki/Expectation%E2%80%93maximization_algorithm) methods. Expectation Maximization (ultra simplified version)(Note: all explanations of EM are far more complicated than seems necessary for our purposes, so here is my overly simplified explanation.)This may make more sense in terms of our earlier Bayesian analyses if we write this as $$p(z=c) = \alpha_k,$$and$$p(x|z=c) = \mathscr{N}(x|\mu_k,\Sigma_k),$$where $z$ is a "hidden" variable related to which "component" each point is assigned to.In the Expectation step, we hold $\mu_k, \Sigma_k$, and $\alpha_k$ fixed and compute the probability that each $x_i$ belongs to component, $c$. In the Maximization step, we hold the probability of the components fixed and maximize $\mu_k, \Sigma_k,$ and $\alpha_k$. Note that $\alpha$ is the relative weight of each Gaussian component and not the probability of each point belonging to a specific component. We can use the following animation to illustrate the process. We start with a 2-component GMM, where the initial components can be randomly determined.The points that are closest to the centroid of a component will be more probable under that distribution in the "E" step and will pull the centroid towards them in the "M" step. Iteration between the "E" and "M" step eventually leads to convergence.In this particular example, 3 components better describes the data and similarly converges. Note that the process is not that sensitive to how the components are first initialized. We pretty much get the same result in the end. | from IPython.display import YouTubeVideo
YouTubeVideo("B36fzChfyGU") | _____no_output_____ | MIT | MixtureModel.ipynb | gtrichards/PHYS_T480 |
A typical call to the [Gaussian Mixture Model](http://scikit-learn.org/stable/modules/mixture.html) algorithm looks like this: | # Execute this cell
import numpy as np
from sklearn.mixture import GMM
X = np.random.normal(size=(1000,2)) #1000 points in 2D
gmm = GMM(3) #three components
gmm.fit(X)
log_dens = gmm.score(X)
BIC = gmm.bic(X) | _____no_output_____ | MIT | MixtureModel.ipynb | gtrichards/PHYS_T480 |
Let's start with the 1-D example given in Ivezic, Figure 6.8, which compares a Mixture Model to KDE.[Note that the version at astroML.org has some bugs!] | # Execute this cell
# Ivezic, Figure 6.8
# Author: Jake VanderPlas
# License: BSD
# The figure produced by this code is published in the textbook
# "Statistics, Data Mining, and Machine Learning in Astronomy" (2013)
# For more information, see http://astroML.github.com
# To report a bug or issue, use the following forum:
# https://groups.google.com/forum/#!forum/astroml-general
%matplotlib inline
import numpy as np
from matplotlib import pyplot as plt
from scipy import stats
from astroML.plotting import hist
from sklearn.mixture import GMM
from sklearn.neighbors import KernelDensity
#------------------------------------------------------------
# Generate our data: a mix of several Cauchy distributions
# this is the same data used in the Bayesian Blocks figure
np.random.seed(0)
N = 10000
mu_gamma_f = [(5, 1.0, 0.1),
(7, 0.5, 0.5),
(9, 0.1, 0.1),
(12, 0.5, 0.2),
(14, 1.0, 0.1)]
true_pdf = lambda x: sum([f * stats.cauchy(mu, gamma).pdf(x)
for (mu, gamma, f) in mu_gamma_f])
x = np.concatenate([stats.cauchy(mu, gamma).rvs(int(f * N))
for (mu, gamma, f) in mu_gamma_f])
np.random.shuffle(x)
x = x[x > -10]
x = x[x < 30]
#------------------------------------------------------------
# plot the results
fig = plt.figure(figsize=(10, 10))
fig.subplots_adjust(bottom=0.08, top=0.95, right=0.95, hspace=0.1)
N_values = (500, 5000)
subplots = (211, 212)
k_values = (10, 100)
for N, k, subplot in zip(N_values, k_values, subplots):
ax = fig.add_subplot(subplot)
xN = x[:N]
t = np.linspace(-10, 30, 1000)
kde = KernelDensity(0.1, kernel='gaussian')
kde.fit(xN[:, None])
dens_kde = np.exp(kde.score_samples(t[:, None]))
# Compute density via Gaussian Mixtures
# we'll try several numbers of clusters
n_components = np.arange(3, 16)
gmms = [GMM(n_components=n).fit(xN[:,None]) for n in n_components]
BICs = [gmm.bic(xN[:,None]) for gmm in gmms]
i_min = np.argmin(BICs)
t = np.linspace(-10, 30, 1000)
logprob, responsibilities = gmms[i_min].score_samples(t[:,None])
# plot the results
ax.plot(t, true_pdf(t), ':', color='black', zorder=3,
label="Generating Distribution")
ax.plot(xN, -0.005 * np.ones(len(xN)), '|k', lw=1.5)
ax.plot(t, np.exp(logprob), '-', color='gray',
label="Mixture Model\n(%i components)" % n_components[i_min])
ax.plot(t, dens_kde, '-', color='black', zorder=3,
label="Kernel Density $(h=0.1)$")
# label the plot
ax.text(0.02, 0.95, "%i points" % N, ha='left', va='top',
transform=ax.transAxes)
ax.set_ylabel('$p(x)$')
ax.legend(loc='upper right')
if subplot == 212:
ax.set_xlabel('$x$')
ax.set_xlim(0, 20)
ax.set_ylim(-0.01, 0.4001)
plt.show() | _____no_output_____ | MIT | MixtureModel.ipynb | gtrichards/PHYS_T480 |
Hmm, that doesn't look so great for the 5000 point distribution. Plot the BIC values and see if anything looks awry. What do the individual components look like? Make a plot of those. Careful with the shapes of the arrays! Can you figure out something that you can do to improve the results? Ivezic, Figure 6.6 shows a 2-D example. In the first panel, we have the raw data. In the second panel we have a density plot (essentially a 2-D histogram). We then try to represent the data with a series of Gaussians. We allow up to 14 Gaussians and use the AIC/BIC to determine the best choice for this number. This is shown in the third panel. Finally, the fourth panel shows the chosen Gaussians with their centroids and 1-$\sigma$ contours.In this case 7 components are required for the best fit. While it looks like we could do a pretty good job with just 2 components, there does appear to be some "background" that is a high enough level to justify further components. | # Execute this cell
# Ivezic, Figure 6.6
# Author: Jake VanderPlas
# License: BSD
# The figure produced by this code is published in the textbook
# "Statistics, Data Mining, and Machine Learning in Astronomy" (2013)
# For more information, see http://astroML.github.com
# To report a bug or issue, use the following forum:
# https://groups.google.com/forum/#!forum/astroml-general
%matplotlib inline
import numpy as np
from matplotlib import pyplot as plt
from scipy.stats import norm
from sklearn.mixture import GMM
from astroML.datasets import fetch_sdss_sspp
from astroML.decorators import pickle_results
from astroML.plotting.tools import draw_ellipse
#------------------------------------------------------------
# Get the Segue Stellar Parameters Pipeline data
data = fetch_sdss_sspp(cleaned=True)
# Note how X was created from two columns of data
X = np.vstack([data['FeH'], data['alphFe']]).T
# truncate dataset for speed
X = X[::5]
#------------------------------------------------------------
# Compute GMM models & AIC/BIC
N = np.arange(1, 14)
#@pickle_results("GMM_metallicity.pkl")
def compute_GMM(N, covariance_type='full', n_iter=1000):
models = [None for n in N]
for i in range(len(N)):
#print N[i]
models[i] = GMM(n_components=N[i], n_iter=n_iter, covariance_type=covariance_type)
models[i].fit(X)
return models
models = compute_GMM(N)
AIC = [m.aic(X) for m in models]
BIC = [m.bic(X) for m in models]
i_best = np.argmin(BIC)
gmm_best = models[i_best]
print "best fit converged:", gmm_best.converged_
print "BIC: n_components = %i" % N[i_best]
#------------------------------------------------------------
# compute 2D density
FeH_bins = 51
alphFe_bins = 51
H, FeH_bins, alphFe_bins = np.histogram2d(data['FeH'], data['alphFe'], (FeH_bins, alphFe_bins))
Xgrid = np.array(map(np.ravel,
np.meshgrid(0.5 * (FeH_bins[:-1]
+ FeH_bins[1:]),
0.5 * (alphFe_bins[:-1]
+ alphFe_bins[1:])))).T
log_dens = gmm_best.score(Xgrid).reshape((51, 51))
#------------------------------------------------------------
# Plot the results
fig = plt.figure(figsize=(12, 5))
fig.subplots_adjust(wspace=0.45, bottom=0.25, top=0.9, left=0.1, right=0.97)
# plot data
ax = fig.add_subplot(141)
ax.scatter(data['FeH'][::10],data['alphFe'][::10],marker=".",color='k',edgecolors='None')
ax.set_xlabel(r'$\rm [Fe/H]$')
ax.set_ylabel(r'$\rm [\alpha/Fe]$')
ax.xaxis.set_major_locator(plt.MultipleLocator(0.3))
ax.set_xlim(-1.101, 0.101)
ax.text(0.93, 0.93, "Input",
va='top', ha='right', transform=ax.transAxes)
# plot density
ax = fig.add_subplot(142)
ax.imshow(H.T, origin='lower', interpolation='nearest', aspect='auto',
extent=[FeH_bins[0], FeH_bins[-1],
alphFe_bins[0], alphFe_bins[-1]],
cmap=plt.cm.binary)
ax.set_xlabel(r'$\rm [Fe/H]$')
ax.set_ylabel(r'$\rm [\alpha/Fe]$')
ax.xaxis.set_major_locator(plt.MultipleLocator(0.3))
ax.set_xlim(-1.101, 0.101)
ax.text(0.93, 0.93, "Density",
va='top', ha='right', transform=ax.transAxes)
# plot AIC/BIC
ax = fig.add_subplot(143)
ax.plot(N, AIC, '-k', label='AIC')
ax.plot(N, BIC, ':k', label='BIC')
ax.legend(loc=1)
ax.set_xlabel('N components')
plt.setp(ax.get_yticklabels(), fontsize=7)
# plot best configurations for AIC and BIC
ax = fig.add_subplot(144)
ax.imshow(np.exp(log_dens),
origin='lower', interpolation='nearest', aspect='auto',
extent=[FeH_bins[0], FeH_bins[-1],
alphFe_bins[0], alphFe_bins[-1]],
cmap=plt.cm.binary)
ax.scatter(gmm_best.means_[:, 0], gmm_best.means_[:, 1], c='w')
for mu, C, w in zip(gmm_best.means_, gmm_best.covars_, gmm_best.weights_):
draw_ellipse(mu, C, scales=[1], ax=ax, fc='none', ec='k')
ax.text(0.93, 0.93, "Converged",
va='top', ha='right', transform=ax.transAxes)
ax.set_xlim(-1.101, 0.101)
ax.set_ylim(alphFe_bins[0], alphFe_bins[-1])
ax.xaxis.set_major_locator(plt.MultipleLocator(0.3))
ax.set_xlabel(r'$\rm [Fe/H]$')
ax.set_ylabel(r'$\rm [\alpha/Fe]$')
plt.show() | _____no_output_____ | MIT | MixtureModel.ipynb | gtrichards/PHYS_T480 |
That said, I'd say that there are *too* many components here. So, I'd be inclined to explore this a bit further if it were my data.Lastly, let's look at a 2-D case where we are using GMM more to characterize the data than to find clusters. | # Execute this cell
# Ivezic, Figure 6.7
# Author: Jake VanderPlas
# License: BSD
# The figure produced by this code is published in the textbook
# "Statistics, Data Mining, and Machine Learning in Astronomy" (2013)
# For more information, see http://astroML.github.com
# To report a bug or issue, use the following forum:
# https://groups.google.com/forum/#!forum/astroml-general
import numpy as np
from matplotlib import pyplot as plt
from sklearn.mixture import GMM
from astroML.datasets import fetch_great_wall
from astroML.decorators import pickle_results
#------------------------------------------------------------
# load great wall data
X = fetch_great_wall()
#------------------------------------------------------------
# Create a function which will save the results to a pickle file
# for large number of clusters, computation will take a long time!
#@pickle_results('great_wall_GMM.pkl')
def compute_GMM(n_clusters, n_iter=1000, min_covar=3, covariance_type='full'):
clf = GMM(n_clusters, covariance_type=covariance_type,
n_iter=n_iter, min_covar=min_covar)
clf.fit(X)
print "converged:", clf.converged_
return clf
#------------------------------------------------------------
# Compute a grid on which to evaluate the result
Nx = 100
Ny = 250
xmin, xmax = (-375, -175)
ymin, ymax = (-300, 200)
Xgrid = np.vstack(map(np.ravel, np.meshgrid(np.linspace(xmin, xmax, Nx),
np.linspace(ymin, ymax, Ny)))).T
#------------------------------------------------------------
# Compute the results
#
# we'll use 100 clusters. In practice, one should cross-validate
# with AIC and BIC to settle on the correct number of clusters.
clf = compute_GMM(n_clusters=100)
log_dens = clf.score(Xgrid).reshape(Ny, Nx)
#------------------------------------------------------------
# Plot the results
fig = plt.figure(figsize=(10, 5))
fig.subplots_adjust(hspace=0, left=0.08, right=0.95, bottom=0.13, top=0.9)
ax = fig.add_subplot(211, aspect='equal')
ax.scatter(X[:, 1], X[:, 0], s=1, lw=0, c='k')
ax.set_xlim(ymin, ymax)
ax.set_ylim(xmin, xmax)
ax.xaxis.set_major_formatter(plt.NullFormatter())
plt.ylabel(r'$x\ {\rm (Mpc)}$')
ax = fig.add_subplot(212, aspect='equal')
ax.imshow(np.exp(log_dens.T), origin='lower', cmap=plt.cm.binary,
extent=[ymin, ymax, xmin, xmax])
ax.set_xlabel(r'$y\ {\rm (Mpc)}$')
ax.set_ylabel(r'$x\ {\rm (Mpc)}$')
plt.show() | _____no_output_____ | MIT | MixtureModel.ipynb | gtrichards/PHYS_T480 |
Initial changes | ds_train = TiggeMRMSDataset(
tigge_dir=f'{DATADRIVE}/tigge/32km/',
tigge_vars=['total_precipitation'],
mrms_dir=f'{DATADRIVE}/mrms/4km/RadarOnly_QPE_06H/',
rq_fn=f'{DATADRIVE}/mrms/4km/RadarQuality.nc',
# const_fn='/datadrive/tigge/32km/constants.nc',
# const_vars=['orog', 'lsm'],
data_period=('2018-01', '2019-01'),
first_days=5,
scale=False
# split='train'
)
mean_precip = []
for idx in range(len(ds_train.idxs)):
X, y = ds_train[idx]
mean_precip.append(y.max())
mean_precip = np.array(mean_precip)
plt.hist(mean_precip, bins=100);
# plt.yscale('log')
cat_bins = np.arange(0, 102, 2, dtype='float')
cat_bins = np.append(np.insert(cat_bins, 1, 0.01), np.inf)
len(cat_bins)
cat_bins
X, y = ds_train[600]
X.shape, y.shape
plt.imshow(y[0])
plt.colorbar();
def to_categorical(y, num_classes=None, dtype='float32'):
"""Copied from keras source code
"""
y = np.array(y, dtype='int')
input_shape = y.shape
if input_shape and input_shape[-1] == 1 and len(input_shape) > 1:
input_shape = tuple(input_shape[:-1])
y = y.ravel()
if not num_classes:
num_classes = np.max(y) + 1
n = y.shape[0]
categorical = np.zeros((n, num_classes), dtype=dtype)
categorical[np.arange(n), y] = 1
output_shape = input_shape + (num_classes,)
categorical = np.reshape(categorical, output_shape)
return categorical
a = pd.cut(y.reshape(-1), cat_bins, labels=False, include_lowest=True).reshape(y.shape)
a.shape
plt.imshow(a[0])
plt.colorbar();
plt.hist(a.flat, bins=cat_bins);
a = to_categorical(a.squeeze(), num_classes=len(cat_bins))
a = np.rollaxis(a, 2)
a.shape | _____no_output_____ | MIT | notebooks/stephan_notebooks/10-Categorical.ipynb | raspstephan/nwp-downscale |
Changes implemented | cat_bins = np.arange(0, 55, 5, dtype='float')
cat_bins = np.append(np.insert(cat_bins, 1, 0.01), np.inf)
len(cat_bins)
plt.plot(cat_bins)
ds_train = TiggeMRMSDataset(
tigge_dir=f'{DATADRIVE}/tigge/32km/',
tigge_vars=['total_precipitation'],
mrms_dir=f'{DATADRIVE}/mrms/4km/RadarOnly_QPE_06H/',
rq_fn=f'{DATADRIVE}/mrms/4km/RadarQuality.nc',
# const_fn='/datadrive/tigge/32km/constants.nc',
# const_vars=['orog', 'lsm'],
data_period=('2018-01', '2018-12'),
first_days=5,
scale=True,
cat_bins=cat_bins
# split='train'
)
ds_valid = TiggeMRMSDataset(
tigge_dir=f'{DATADRIVE}/tigge/32km/',
tigge_vars=['total_precipitation'],
mrms_dir=f'{DATADRIVE}/mrms/4km/RadarOnly_QPE_06H/',
rq_fn=f'{DATADRIVE}/mrms/4km/RadarQuality.nc',
# const_fn='/datadrive/tigge/32km/constants.nc',
# const_vars=['orog', 'lsm'],
data_period=('2019-01', '2019-12'),
first_days=2,
scale=True,
cat_bins=cat_bins,
mins=ds_train.mins,
maxs=ds_train.maxs
# split='train'
)
X, y = ds_train[600]
X.shape, y.shape
y
sampler_train = torch.utils.data.WeightedRandomSampler(ds_train.compute_weights(), len(ds_train), replacement=True)
sampler_valid = torch.utils.data.WeightedRandomSampler(ds_valid.compute_weights(), len(ds_valid), replacement=True)
dl_train = torch.utils.data.DataLoader(ds_train, batch_size=32, sampler=sampler_train)
dl_valid = torch.utils.data.DataLoader(ds_valid, batch_size=32, sampler=sampler_valid)
len(dl_train)
len(ds_train)
X, y = next(iter(dl_train))
fig, axs = plt.subplots(4, 8, figsize=(24, 12))
for x, ax in zip(X.numpy(), axs.flat):
ax.imshow(x[0], cmap='gist_ncar_r', vmin=0, vmax=0.5)
X.shape, y.shape
y.type()
class UpsampleBlock(nn.Module):
def __init__(self, nf, spectral_norm=False, method='PixelShuffle'):
super().__init__()
self.conv = nn.Conv2d(nf, nf * 4 if method=='PixelShuffle' else nf, kernel_size=3, stride=1, padding=1)
if method == 'PixelShuffle':
self.upsample = nn.PixelShuffle(2)
elif method == 'bilinear':
self.upsample = nn.Upsample(scale_factor=2, mode='bilinear')
else:
raise NotImplementedError
self.activation = nn.LeakyReLU(0.2)
if spectral_norm:
self.conv = nn.utils.spectral_norm(self.conv)
def forward(self, x):
out = self.conv(x)
out = self.upsample(out)
out = self.activation(out)
return out
class Generator(nn.Module):
"""Generator with noise vector and spectral normalization """
def __init__(self, nres, nf_in, nf, relu_out=False, use_noise=True, spectral_norm=True,
nout=1, softmax_out=False, upsample_method='PixelShuffle'):
""" General Generator with different options to use. e.g noise, Spectral normalization (SN) """
super().__init__()
self.relu_out = relu_out
self.softmax_out = softmax_out
self.use_noise = use_noise
self.spectral_norm = spectral_norm
# First convolution
if use_noise:
self.conv_in = nn.Conv2d(nf_in, nf-1, kernel_size=9, stride=1, padding=4)
else:
self.conv_in = nn.Conv2d(nf_in, nf, kernel_size=9, stride=1, padding=4)
self.activation_in = nn.LeakyReLU(0.2)
# Resblocks keeping shape
self.resblocks = nn.Sequential(*[
ResidualBlock(nf, spectral_norm=spectral_norm) for i in range(nres)
])
# Resblocks with upscaling
self.upblocks = nn.Sequential(*[
UpsampleBlock(nf, spectral_norm=spectral_norm, method=upsample_method) for i in range(3)
])
self.conv_out = nn.Conv2d(nf, nout, kernel_size=9, stride=1, padding=4)
if spectral_norm:
self.conv_in = nn.utils.spectral_norm(self.conv_in)
self.conv_out = nn.utils.spectral_norm(self.conv_out)
def forward(self, x):
out = self.conv_in(x)
out = self.activation_in(out)
if self.use_noise:
bs, _, h, w = x.shape
z = torch.normal(0, 1, size=(bs, 1, h, w), device=device, requires_grad=True)
out = torch.cat([out, z], dim=1)
skip = out
out = self.resblocks(out)
out = out + skip
out = self.upblocks(out)
out = self.conv_out(out)
if self.relu_out:
out = nn.functional.relu(out)
if self.softmax_out:
out = nn.functional.softmax(out, dim=1)
return out
gen = Generator(
nres=3, nf_in=1, nf=64, relu_out=False, use_noise=False, spectral_norm=False,
nout=len(cat_bins)-1, softmax_out=False, upsample_method='bilinear'
).to(device)
count_parameters(gen)
criterion = nn.CrossEntropyLoss()
optimizer = torch.optim.Adam(gen.parameters(), lr=1e-4)
trainer = Trainer(gen, optimizer, criterion, dl_train, dl_valid)
trainer.fit(10)
trainer.plot_losses()
preds = nn.functional.softmax(gen(X.to(device)), dim=1).cpu().detach().numpy()
preds.shape
target = y.cpu().detach().numpy()
target.shape
np.argmax(target.mean((1, 2)))
i=7
plt.imshow(target[i])
plt.colorbar()
target[i, 20, 20]
plt.imshow(X[i, 0] * ds_train.maxs.tp.values)
plt.colorbar();
plt.imshow(np.argmax(preds[i], axis=0))
plt.colorbar()
plt.plot(preds[i, :, 100, 100])
plt.axvline(target[i, 100, 100])
cdf = np.cumsum(preds, axis=1)
cdf.shape
len(cat_bins)
plt.plot(cat_bins[:-1], cdf[i, :, 20, 20])
plt.plot(cat_bins[:-1], cdf[i, :, 100, 100])
from scipy.ndimage import gaussian_filter
def corr_random2D(size, sigma=5, inflation=0.5):
r = np.random.uniform(size=(size, size))
r = gaussian_filter(r, sigma)
r = (r - 0.5) * (inflation / r.std()) + 0.5
r = 1/(1 + np.exp(-r))
return r
rand = corr_random2D(128, 3)
plt.imshow(rand)
plt.colorbar();
cat_bins
c = cdf[i, :, 100, 100]
c = np.insert(c, 0, 0)
c
len(cat_bins), len(c)
p = rand[100, 100]
p
b = np.digitize(p, c, right=True) - 1
b
c[b], c[b+1]
w1 = (p - c[b]) / (c[b+1] - c[b])
w2 = (c[b+1] - p) / (c[b+1] - c[b])
w1, w2
v = cat_bins[b] * w1 + cat_bins[b+1] * w2
v
def cat2real1D(pdf, q, cat_bins, interpolate=True):
c = np.cumsum(pdf)
c = np.insert(c, 0, 0)
b = np.digitize(q, c, right=True) - 1
if interpolate:
w1 = (q - c[b]) / (c[b+1] - c[b])
w2 = (c[b+1] - q) / (c[b+1] - c[b])
else:
w1, w2 = 0.5, 0.5
assert w1 >0, 'Weights must be positive'
assert w2 >0, 'Weights must be positive'
v = cat_bins[b] * w2 + cat_bins[b+1] * w1
return v
import pdb
def cat2real2D(pdf, q, cat_bins, interpolate=True):
nbins, nx, ny = pdf.shape
# pdb.set_trace()
r = [cat2real1D(a, b, cat_bins, interpolate) for a, b in zip(pdf.reshape(nbins, -1).T, q.reshape(-1))]
r = np.array(r).reshape(nx, ny)
return r
o = cat2real2D(pdf, rand, cat_bins)
plt.imshow(o)
plt.colorbar();
weights = ds_valid.compute_weights()
np.argsort(weights)[::-1][:20]
X_sample, y_sample = ds_valid.__getitem__(501, no_cat=True)
X_sample.shape, y_sample.shape
vmin=0
vmax=10
fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(10, 5))
img = ax1.imshow(X_sample[0]*ds_valid.maxs.tp.values, cmap='gist_ncar_r', vmin=vmin, vmax=vmax)
plt.colorbar(img, ax=ax1, shrink=0.7)
img = ax2.imshow(y_sample[0], cmap='gist_ncar_r', vmin=vmin, vmax=vmax)
plt.colorbar(img, ax=ax2, shrink=0.7)
pdf = nn.functional.softmax(gen(torch.from_numpy(X_sample[None]).to(device))).cpu().detach().numpy()[0]
pdf.shape
fig, axs = plt.subplots(1, 5, figsize=(20, 4))
for i in range(5):
if i ==0:
rand = np.ones((128, 128)) * 0.5
else:
rand = corr_random2D(128, 3, inflation=0.3)
o = cat2real2D(pdf, rand, cat_bins)
axs[i].imshow(o, cmap='gist_ncar_r', vmin=vmin, vmax=vmax) | _____no_output_____ | MIT | notebooks/stephan_notebooks/10-Categorical.ipynb | raspstephan/nwp-downscale |
Overfitting test MSE | ds_train = TiggeMRMSDataset(
tigge_dir=f'{DATADRIVE}/tigge/32km/',
tigge_vars=['total_precipitation'],
mrms_dir=f'{DATADRIVE}/mrms/4km/RadarOnly_QPE_06H/',
rq_fn=f'{DATADRIVE}/mrms/4km/RadarQuality.nc',
# const_fn='/datadrive/tigge/32km/constants.nc',
# const_vars=['orog', 'lsm'],
data_period=('2018-01', '2018-12'),
first_days=5,
scale=True,
# cat_bins=cat_bins
# split='train'
)
ds_valid = TiggeMRMSDataset(
tigge_dir=f'{DATADRIVE}/tigge/32km/',
tigge_vars=['total_precipitation'],
mrms_dir=f'{DATADRIVE}/mrms/4km/RadarOnly_QPE_06H/',
rq_fn=f'{DATADRIVE}/mrms/4km/RadarQuality.nc',
# const_fn='/datadrive/tigge/32km/constants.nc',
# const_vars=['orog', 'lsm'],
data_period=('2019-01', '2019-12'),
first_days=2,
scale=True,
# cat_bins=cat_bins,
mins=ds_train.mins,
maxs=ds_train.maxs
# split='train'
)
X, y = ds_train[600]
X.shape, y.shape
sampler_train = torch.utils.data.WeightedRandomSampler(ds_train.compute_weights(), len(ds_train))
sampler_valid = torch.utils.data.WeightedRandomSampler(ds_valid.compute_weights(), len(ds_valid))
dl_train = torch.utils.data.DataLoader(ds_train, batch_size=32, sampler=sampler_train)
dl_valid = torch.utils.data.DataLoader(ds_valid, batch_size=32, sampler=sampler_valid)
len(dl_train)
len(ds_train)
X, y = next(iter(dl_train))
fig, axs = plt.subplots(4, 8, figsize=(24, 12))
for x, ax in zip(X.numpy(), axs.flat):
ax.imshow(x[0], cmap='gist_ncar_r', vmin=0, vmax=0.5)
X.shape, y.shape
y.type()
class UpsampleBlock(nn.Module):
def __init__(self, nf, spectral_norm=False, method='PixelShuffle'):
super().__init__()
self.conv = nn.Conv2d(nf, nf * 4 if method=='PixelShuffle' else nf, kernel_size=3, stride=1, padding=1)
if method == 'PixelShuffle':
self.upsample = nn.PixelShuffle(2)
elif method == 'bilinear':
self.upsample = nn.Upsample(scale_factor=2, mode='bilinear')
else:
raise NotImplementedError
self.activation = nn.LeakyReLU(0.2)
if spectral_norm:
self.conv = nn.utils.spectral_norm(self.conv)
def forward(self, x):
out = self.conv(x)
out = self.upsample(out)
out = self.activation(out)
return out
class Generator(nn.Module):
"""Generator with noise vector and spectral normalization """
def __init__(self, nres, nf_in, nf, relu_out=False, use_noise=True, spectral_norm=True,
nout=1, softmax_out=False, upsample_method='PixelShuffle'):
""" General Generator with different options to use. e.g noise, Spectral normalization (SN) """
super().__init__()
self.relu_out = relu_out
self.softmax_out = softmax_out
self.use_noise = use_noise
self.spectral_norm = spectral_norm
# First convolution
if use_noise:
self.conv_in = nn.Conv2d(nf_in, nf-1, kernel_size=9, stride=1, padding=4)
else:
self.conv_in = nn.Conv2d(nf_in, nf, kernel_size=9, stride=1, padding=4)
self.activation_in = nn.LeakyReLU(0.2)
# Resblocks keeping shape
self.resblocks = nn.Sequential(*[
ResidualBlock(nf, spectral_norm=spectral_norm) for i in range(nres)
])
# Resblocks with upscaling
self.upblocks = nn.Sequential(*[
UpsampleBlock(nf, spectral_norm=spectral_norm, method=upsample_method) for i in range(3)
])
self.conv_out = nn.Conv2d(nf, nout, kernel_size=9, stride=1, padding=4)
if spectral_norm:
self.conv_in = nn.utils.spectral_norm(self.conv_in)
self.conv_out = nn.utils.spectral_norm(self.conv_out)
def forward(self, x):
out = self.conv_in(x)
out = self.activation_in(out)
if self.use_noise:
bs, _, h, w = x.shape
z = torch.normal(0, 1, size=(bs, 1, h, w), device=device, requires_grad=True)
out = torch.cat([out, z], dim=1)
skip = out
out = self.resblocks(out)
out = out + skip
out = self.upblocks(out)
out = self.conv_out(out)
if self.relu_out:
out = nn.functional.relu(out)
if self.softmax_out:
out = nn.functional.softmax(out, dim=1)
return out
gen = Generator(
nres=3, nf_in=1, nf=64, relu_out=False, use_noise=False, spectral_norm=False,
nout=1, softmax_out=False, upsample_method='PixelShuffle'
).to(device)
count_parameters(gen)
criterion = nn.MSELoss()
optimizer = torch.optim.Adam(gen.parameters(), lr=1e-5)
trainer = Trainer(gen, optimizer, criterion, dl_train, dl_valid)
trainer.fit(10)
trainer.plot_losses()
plot_sample(X, y, gen, 13) | _____no_output_____ | MIT | notebooks/stephan_notebooks/10-Categorical.ipynb | raspstephan/nwp-downscale |
--- Saving your plot to a file (or to an io buffer):- `mplfinance.plot()` allows you to save your plot to a file, or io-buffer, using the `savefig` keyword.- The value of `savefig` may be a `str`, `dict`, or `io.ByteIO` object. - If the value is a `str` then it is assumed to be the file name to which to save the figure/plot. - If the value is an `io.ByteIO` object, then the figure will be saved to the io buffer object. This avoids interaction with disk, and can also be useful when mplfinance is behind a web server (so that requests for an image file can be serviced without going to disk).If the file extension is one of those recognized by `matplotlib.pyplot.savefig()` then the file type will be inferred from the extension, for example: `.pdf`, `.svg`, `.png`, `.jpg` ... | df = pd.read_csv('data/SP500_NOV2019_Hist.csv',index_col=0,parse_dates=True)
%%capture
## cell magic function `%%capture` blocks jupyter notebook output,
## which is not needed here since the plot is saved to a file anyway:
mpf.plot(df,type='candle',volume=True,savefig='testsave.png') | _____no_output_____ | Apache-2.0 | examples/savefig.ipynb | fadeawaylove/stock-trade-system |
--- We can use IPython.display.Image to display the image file here in the notebook: | import IPython.display as IPydisplay
%ls -l testsave.png
IPydisplay.Image(filename='testsave.png') | -rw-rw-rw- 1 dino dino 24877 Jun 7 17:52 [0m[01;35mtestsave.png[0m
| Apache-2.0 | examples/savefig.ipynb | fadeawaylove/stock-trade-system |
--- We can use io to save the plot as a byte buffer: | %%capture
## cell magic function `%%capture` blocks jupyter notebook output,
## which is not needed here, since the plot is saved to the io-buffer anyway:
buf = io.BytesIO()
mpf.plot(df,type='candle',volume=True,savefig=buf)
buf.seek(0) | _____no_output_____ | Apache-2.0 | examples/savefig.ipynb | fadeawaylove/stock-trade-system |
We can use Ipython.display.Image to display the image in the ioBytes buffer: | IPydisplay.Image(buf.read()) | _____no_output_____ | Apache-2.0 | examples/savefig.ipynb | fadeawaylove/stock-trade-system |
--- Specifying image attributes with `savefig`We can control various attributes of the saved figure/plot by passing a `dict`ionary as the value for the `savefig` keyword.The dictionary **must** contain the keyword `fname` for the file name to be saved, **and *may* contain any of the other keywords accepted by [`matplotlib.pyplot.savefig()`](https://matplotlib.org/3.1.1/api/_as_gen/matplotlib.pyplot.savefig.html)** (for example: dpi, facecolor, edgecolor, orientation, format, metadata, quality)When creating the `dict`, I recommend using the `dict()` constructor so that that `keyword=` syntax may be used and thereby more closely resemble calling:**[`matplotlib.pyplot.savefig()`](https://matplotlib.org/3.1.1/api/_as_gen/matplotlib.pyplot.savefig.html)** | %%capture
## %%capture blocks jupyter notebook output; plots are saved to files anyway:
save = dict(fname='tsave30.jpg',dpi=30,pad_inches=0.25)
mpf.plot(df,volume=True,savefig=save)
mpf.plot(df,volume=True,savefig=dict(fname='tsave100.jpg',dpi=100,pad_inches=0.25))
%ls -l tsave30.jpg
%ls -l tsave100.jpg
IPydisplay.Image(filename='tsave30.jpg')
IPydisplay.Image(filename='tsave100.jpg') | -rw-rw-rw- 1 dino dino 11016 Jun 7 17:52 [0m[01;35mtsave30.jpg[0m
-rw-rw-rw- 1 dino dino 54172 Jun 7 17:52 [0m[01;35mtsave100.jpg[0m
| Apache-2.0 | examples/savefig.ipynb | fadeawaylove/stock-trade-system |
Specifying image attributes (via `savefig`) dict also works with an io.BytesIO buffer:- Just assign the io-buffer to the `fname` key in the savefig dict | %%capture
buf30dpi = io.BytesIO()
buf100dpi = io.BytesIO()
mpf.plot(df,volume=True,savefig=dict(fname=buf30dpi ,dpi=30 ,pad_inches=0.25))
mpf.plot(df,volume=True,savefig=dict(fname=buf100dpi,dpi=100,pad_inches=0.25)) | _____no_output_____ | Apache-2.0 | examples/savefig.ipynb | fadeawaylove/stock-trade-system |
Use Ipython.display.Image to display the buffer contents: | _ = buf30dpi.seek(0)
IPydisplay.Image(buf30dpi.read())
_ = buf100dpi.seek(0)
IPydisplay.Image(buf100dpi.read()) | _____no_output_____ | Apache-2.0 | examples/savefig.ipynb | fadeawaylove/stock-trade-system |
Oxford University COVID-19 forecasting project dataThe data below was downloaded from the [Oxford University countermeasures database](https://www.notion.so/COVID-19-countermeasures-database-b532a58d6f944ef6982ab565627bdb08). See the website for a description of the data sources. | countermeasures_df = pd.read_csv("data/containment_measures_march23.csv")
countermeasures_df.head(1)
countermeasures_df[["Country", "Date Start", "Description of measure implemented"]].groupby("Country").head(10)
print(countermeasures_df["Country"].unique())
print(countermeasures_df["Implementing City"].unique())
print(countermeasures_df["Implementing State/Province"].unique()) | [nan 'Quang Ninh' 'Gyeongbook Province' 'Daegu, Gyeongbook Province'
'Busan' 'Hubei' 'Hunan' 'Tianjin' 'Zheijang' 'Meituan' 'Chongqing'
'Zhejian' 'Shanghai' 'Sichuan' 'Jiangsu' 'Guangdong' 'Hangzhou'
'Shenzhen' 'Leishenshan' 'Guangzhou' 'Kanagawa prefecture' 'Guangxi'
'Guizhou' 'Huanggang' 'Henan, Shandong' 'Liaoning, Shandong' 'Shandong'
'Madrid' 'Tyrol' 'Basque Country' 'Galicia' 'Republika Srpska' 'Bosnia'
'Autonomous Region of Madeira' 'California' 'Leningrad Oblast'
'Moscow Oblast' 'St. Petersburg Oblast' 'Santa Clara County' 'Seattle'
'Berkley, Countra Costa Country, Santa Clara County, los Angles'
'Orange County' 'Placer County, San Mateo County, Sonoma County'
'San Benito County, Santa Clara County'
'Kershan County, Lancaster County' 'Bavaria' 'Colima' 'Mexico City'
'South Fulton' 'Atalanta, Brookhaven, Clarkston, Dunwoody' 'Atalanta'
'Albany, Athens-Clark county, Bosnia, Dougherty County'
'Gyeonggi-Province, Paju' 'Gyeonggi-Province, Seoul'
'Farifax county, Loudoun county, Prince William County, Stafford County']
| MIT | notebooks/data_sources.ipynb | braadbaart/covid19 |
John Hopkins containment measures databaseThe data is made available as part of the John Hopkins [Containment Measures Database](http://epidemicforecasting.org/containment). See the website in the link for a description of the data sources. | containment_df = pd.read_csv("data/countermeasures_db_johnshopkins_2020_03_30.csv")
containment_df.columns
print(containment_df["Country"].unique())
cases_df = containment_df[["Date", "Country", "Confirmed Cases", "Deaths"]]\
.loc[containment_df["Confirmed Cases"] > 3000]\
.pivot(index="Date", columns="Country", values="Confirmed Cases")
cases_df.plot(figsize=(16,8), title="Per-country growth in confirmed cases after the first 3000")\
.legend(bbox_to_anchor=(1,1))
deaths_df = containment_df[["Date", "Country", "Confirmed Cases", "Deaths"]]\
.loc[containment_df["Deaths"] > 100]\
.pivot(index="Date", columns="Country", values="Deaths")
deaths_df.plot(figsize=(16,8), title="Deaths per country after the first 100 deaths recorded")\
.legend(bbox_to_anchor=(1,1))
other_cm_cols = ['Unnamed: 0', 'Resumption', 'Diagnostic criteria loosened', 'Testing criteria', 'Date', 'Country',
'Confirmed Cases', 'Deaths']
countermeasures = list(filter(lambda m: m not in other_cm_cols, containment_df.columns))
cm_df = containment_df[countermeasures + ['Date', 'Country']].fillna(0)
cm_df[countermeasures] = cm_df[countermeasures].mask(cm_df[countermeasures] > 0, 1)
cm_df.groupby("Date").sum().plot(figsize=(16,8), title="Number of countries implementing measure by date")\
.legend(bbox_to_anchor=(1,1)) | _____no_output_____ | MIT | notebooks/data_sources.ipynb | braadbaart/covid19 |
**Rendering component declaration.** | # imports for setting up display for the colab server.
!sudo apt-get update > /dev/null 2>&1
!sudo apt-get install -y xvfb x11-utils > /dev/null 2>&1
!pip install gym==0.17.* pyvirtualdisplay==0.2.* PyOpenGL==3.1.* PyOpenGL-accelerate==3.1.* > /dev/null 2>&1
# gym related import statements.
import gym
from gym import logger as gymlogger
from gym.wrappers import Monitor
gymlogger.set_level(40) #error only
# RL agent construction related imports.
import numpy as np
np.random.seed(0)
import matplotlib.pyplot as plt
from scipy.special import softmax
# virtual display related import statements.
import math
import glob
import io
import base64
import time
from time import sleep
from tqdm import tqdm
from IPython.display import HTML
from IPython import display as ipythondisplay
# This creates virtual display to send the frames for being rendered.
from pyvirtualdisplay import Display
display = Display(visible=0, size=(1366, 768))
display.start()
def show_video():
'''
This function loads the data video inline into the colab notebook.
By reading the video stored by the Monitor class.
'''
mp4list = glob.glob('video/*.mp4')
if len(mp4list) > 0:
mp4 = mp4list[0]
video = io.open(mp4, 'r+b').read()
encoded = base64.b64encode(video)
ipythondisplay.display(HTML(data='''<video alt="test" autoplay
loop controls style="height: 400px;">
<source src="data:video/mp4;base64,{0}" type="video/mp4" />
</video>'''.format(encoded.decode('ascii'))))
else:
print("Could not find video")
def wrap_env(env):
'''
This monitoring tool records the outputs from the output and saves it a
mp4 file in the stated directory. If we don't change the video directory
the videos will get stored in 'content/' directory.
'''
env = Monitor(env, './video', force=True)
return env | _____no_output_____ | Unlicense | milestone-two/sarsa_lambda_agent_mountain_car_v0.ipynb | galleon/prototyping-self-driving-agents |
**SARSA(Lambda) Algorithm Implementation for MountainCarV0 Environment**__The implementation consists of below stated sections:__ * __Agent class decleration and parsing environment.__* __SARSA(LAMBDA) Algorithm Implementation.__* __Plotting results, outputting results and downloading them.__ **Agent Class Decleration and Parsing Environment** | # This environment has two degrees of freedom: position and velocity.
# Our agent will learn to interact with environment having these two values.
class State:
def __init__(self):
self.pos = None
self.vel = None
# Agent class defined for storing all the agent related values.
# and getting actions from the policy. Here, target policy is same as behavior policy.
class Agent:
def __init__(self, env):
self.velocity_lim = np.array([env.observation_space.low[1], env.observation_space.high[1]])
self.position_lim = np.array([env.observation_space.low[0], env.observation_space.high[0]])
self.velocity_step, self.position_step = 0.005, 0.1
self.velocity_space = np.arange(self.velocity_lim[0], self.velocity_lim[1]
+ self.velocity_step, self.velocity_step)
self.position_space = np.arange(self.position_lim[0], self.position_lim[1]
+ self.position_step, self.position_step)
self.m, self.n, self.n_action = len(self.velocity_space), len(self.position_space), 3
self.Q_sa = np.full(shape = (self.m, self.n, 3),
fill_value = 0.0, dtype = np.float32)
self.collective_record = []
self.success = []
def get_action_value_index(self, state):
pos_offset = state[0] - self.position_lim[0]
vel_offset = state[1] - self.velocity_lim[0]
pos_ind = pos_offset // self.position_step
vel_ind = vel_offset // self.velocity_step
return np.array([vel_ind, pos_ind], dtype= np.int)
def get_action(self, state):
ind = self.get_action_value_index(state, 0)
p = self.Policy[ind[0], ind[1], :]
action = np.random.choice([0, 1, 2], size = 1, p = p)
return action[0]
# Wraping the environment in the Monitor class.
env = wrap_env(gym.make('MountainCar-v0'))
# Fixing the randomness in the environment.
env.seed(0)
# Parsing the environment onto Agent object.
sarsa_agent = Agent(env)
# For some information into Q(s,a) table generated.
# It's dimension is equal to A(Q(s,a)) = n[D(1)]*n[D(2)]*...*n[D(f)]*n(D(actions))
print("Q Shape = ",sarsa_agent.Q_sa.shape) | Q Shape = (30, 20, 3)
| Unlicense | milestone-two/sarsa_lambda_agent_mountain_car_v0.ipynb | galleon/prototyping-self-driving-agents |
**SARSA(Lambda) algorithm implementation.** | # SARSA(lambda) algorithm takes part from TD(0) and TD(1) algorithm.
eps = 0.8 # greedy epsilon exploration-vs-exploitation variable.
changed_eps= []
changes_alpha= []
alpha = 0.2 # learning rate value
lambda_val = 0.8 # credit assignment variable to previous states.
alpha_decay = 0.999
eps_decay = 0.995
sarsa_agent.e = np.zeros(shape = (sarsa_agent.m, sarsa_agent.n, 3)) # eligibility of all states.
finish = False
num_iter = 2000
for i_eps in tqdm(range(1, num_iter + 1)):
state = env.reset()
sarsa_agent.e[:, :, :] = 0
gamma = 1.0
ind = sarsa_agent.get_action_value_index(state)
# greedy exploration and exploitation step.
if np.random.random() < 1 - eps:
action = np.argmax(sarsa_agent.Q_sa[ind[0], ind[1], :])
else:
action = np.random.randint(0, 3)
# running episodes for 200 times for this environment.
for t in range(201):
ind = sarsa_agent.get_action_value_index(state)
next_state, reward, done, info = env.step(action)
next_ind = sarsa_agent.get_action_value_index(next_state)
if np.random.random() < 1 - eps:
next_action = np.argmax(sarsa_agent.Q_sa[next_ind[0], next_ind[1], :])
else:
next_action = np.random.randint(0, 3)
# forward view T(lambda) SARSA equation for making updates.
delta = reward + gamma * sarsa_agent.Q_sa[next_ind[0],next_ind[1], next_action] - sarsa_agent.Q_sa[ind[0],ind[1],action]
sarsa_agent.e[ind[0],ind[1],action] += 1
sarsa_agent.Q_sa = np.add(sarsa_agent.Q_sa, np.multiply(alpha * delta, sarsa_agent.e))
sarsa_agent.e = np.multiply(gamma * lambda_val, sarsa_agent.e)
if done:
if t < 199:
sarsa_agent.success.append((i_eps, t))
sarsa_agent.collective_record.append(-t)
eps = max(0.0, eps * eps_decay)
alpha = max(0.0, alpha * alpha_decay)
break
state = next_state
action = next_action | 100%|██████████| 2000/2000 [00:52<00:00, 37.97it/s]
| Unlicense | milestone-two/sarsa_lambda_agent_mountain_car_v0.ipynb | galleon/prototyping-self-driving-agents |
**Plotting reward fuction resuls and displaying output.** | # This graph shows the saturation of rewards to a minimum under 1000 episodes.
fig, ax = plt.subplots(figsize = (9, 5))
plt.plot(sarsa_agent.collective_record[:],'.')
plt.yticks(range(-110, -200, -10))
plt.ylabel("reward function values")
plt.xlabel("episode number count")
plt.grid()
plt.show()
# Calculating the mean performance of the agent.
for i_eps in (range(1, 100)):
state = env.reset()
gamma = 1.0
ind = sarsa_agent.get_action_value_index(state)
action = np.argmax(sarsa_agent.Q_sa[ind[0], ind[1], :])
for t in range(201):
env.render()
ind = sarsa_agent.get_action_value_index(state)
next_state, reward, done, info = env.step(action)
next_ind = sarsa_agent.get_action_value_index(next_state)
next_action = np.argmax(sarsa_agent.Q_sa[next_ind[0], next_ind[1], :])
if done:
if t < 199:
sarsa_agent.success.append((i_eps, t))
sarsa_agent.collective_record.append(-t)
sleep(1)
break
state = next_state
action = next_action
# Plotting the mean performance of the agent.
fig, ax = plt.subplots(figsize = (9, 5))
plt.plot(sarsa_agent.collective_record[-100:], '-')
plt.yticks(range(-110, -200, -10))
plt.title("Test Results")
plt.ylabel("Mean reward function value")
plt.xlabel("Episode number count")
plt.grid()
plt.show()
# Demonstrating the output of the agent's working.
show_video()
# zipping the video folder for the given SARSA agent.
!zip -r /content/file.zip /content/video
# downloading the file resource.
from google.colab import files
files.download("/content/file.zip")
| _____no_output_____ | Unlicense | milestone-two/sarsa_lambda_agent_mountain_car_v0.ipynb | galleon/prototyping-self-driving-agents |
Data | df = pd.read_csv("HW5_data.csv")
x = df[["X", "Y"]].values
y = df.Z.values
df.head()
df.describe() | _____no_output_____ | MIT | .ipynb_checkpoints/polynomial_feat-checkpoint.ipynb | borab96/misc-notebooks |
The data is split into training, validation and test sets with ratios $0.6:0.2:0.2$ | x_train, x_val, x_test, y_train, y_val, y_test = train_test_val_split(x, y)
print("Training size")
print(x_train.shape)
print("Validation size")
print(x_val.shape)
print("Test size")
print(x_test.shape) | Training size
(600, 2)
Validation size
(200, 2)
Test size
(200, 2)
| MIT | .ipynb_checkpoints/polynomial_feat-checkpoint.ipynb | borab96/misc-notebooks |
Model
We are told to apply a polynomial transformation on the features and use linear regression to obtain the optimal coefficients of the polynomial fit. We create a simple pipeline to achieve this. The only hyperparameter is the maximal degree of the polynomial feature map. The feature map ignores the bias but the linear regressor does not which correctly accounts for the data not being centered.
In order to tune the degree hyperparameter we sweep the parameter space and compute the MSE on the unseen validation set we had created. I am assuming the question does not require anything beyond this simple approach. | def poly_fit_pipeline(degree):
polynomial_features = PolynomialFeatures(degree=degree, include_bias=False)
pipeline = Pipeline([("polynomial_features", polynomial_features), ("linear_regression", LinearRegression())])
return pipeline
degrees = [2,3,4,5,6,7]
mse = []
for d in degrees:
model = poly_fit_pipeline(d)
model.fit(x_train, y_train)
y_pred = model.predict(x_val)
mse.append(mean_squared_error(y_pred, y_val))
print("Lowest validation MSE: "+str(round(np.min(mse),3)))
print("Optimal degree: "+str(degrees[np.argmin(mse)]))
plt.figure()
plt.plot(degrees, np.log(mse))
plt.ylabel(r"$\log MSE$")
plt.xlabel("D")
model_optimal = poly_fit_pipeline(6)
model_optimal.fit(x_train, y_train)
mse_train = mean_squared_error(y_train, model.predict(x_train))
print("Training MSE: "+str(mse_train))
mse_val = mean_squared_error(y_val, model.predict(x_val))
print("Validation MSE: "+str(mse_val))
mse_test = mean_squared_error(y_test, model.predict(x_test))
print("Test MSE: "+str(mse_test) ) | Training MSE: 0.008621014128936854
Validation MSE: 0.00895723100171667
Test MSE: 0.009421933878894272
| MIT | .ipynb_checkpoints/polynomial_feat-checkpoint.ipynb | borab96/misc-notebooks |
The optimization over the hyperparameter space indicates that a maximal degree of $6$ is optimal. Notice that the difference between a degree 5 fit and a degree 6 fit is miniscule so one could also go with the less complex model that offers very similar accuracy. We'll stick with $D=6$ though.
Details of the optimal model | print("Model parameters")
print(model_optimal.get_params())
print("regression coefficients")
print(model_optimal['linear_regression'].coef_)
print("regression intercept")
print(model_optimal['linear_regression'].intercept_)
| Model parameters
{'memory': None, 'steps': [('polynomial_features', PolynomialFeatures(degree=6, include_bias=False, interaction_only=False,
order='C')), ('linear_regression', LinearRegression(copy_X=True, fit_intercept=True, n_jobs=None, normalize=False))], 'verbose': False, 'polynomial_features': PolynomialFeatures(degree=6, include_bias=False, interaction_only=False,
order='C'), 'linear_regression': LinearRegression(copy_X=True, fit_intercept=True, n_jobs=None, normalize=False), 'polynomial_features__degree': 6, 'polynomial_features__include_bias': False, 'polynomial_features__interaction_only': False, 'polynomial_features__order': 'C', 'linear_regression__copy_X': True, 'linear_regression__fit_intercept': True, 'linear_regression__n_jobs': None, 'linear_regression__normalize': False}
regression coefficients
[ 8.50234361e-01 9.62565178e-01 2.44544103e-01 -2.94295566e+00
5.06385541e-03 2.24830034e-03 4.74477640e-01 9.21766117e-04
-1.47698485e-03 2.97942590e-01 2.64905776e-03 2.57614722e-03
-1.42506563e-03 2.57835307e-04 2.44146426e-04 -6.50061514e-07
-4.57242498e-04 2.01383539e-04 4.08693006e-06 1.20000278e+00
-9.28090109e-06 -4.25226714e-06 9.58005035e-06 1.34097259e-05
-2.02810867e-05 7.75568090e-06 -1.56872787e-06]
regression intercept
5.092513526440598
| MIT | .ipynb_checkpoints/polynomial_feat-checkpoint.ipynb | borab96/misc-notebooks |
Object Detection Setup | #@title
import os
!pip install --quiet tensorflow_text
os.environ["TFHUB_MODEL_LOAD_FORMAT"] = "COMPRESSED"
#@title
import os
import math
import numpy as np
import requests
import tensorflow as tf
import tensorflow_hub as hub
import tensorflow_text as tf_text
import tensorflow_datasets as tfds
import matplotlib.pyplot as plt
import seaborn as sns
sns.set_style("whitegrid", {'axes.grid' : False})
%load_ext tensorboard
import requests
def download_image(url, path):
r = requests.get(url, allow_redirects=True)
with open(path, 'wb') as f:
f.write(r.content)
return path
def plot(y, titles=None):
for i, image in enumerate(y):
if image is None:
plt.subplot(1, len(y), i+1)
plt.axis('off')
continue
t = titles[i] if titles else None
plt.subplot(1, len(y), i+1, title=t)
plt.imshow(image)
plt.axis('off')
plt.tight_layout() | _____no_output_____ | Apache-2.0 | notebooks/supervised/detection/object-detection-efficientdet-d4.ipynb | lucasdavid/algorithms-in-tensorflow |
Model Definition | detector = hub.load("https://tfhub.dev/tensorflow/efficientdet/d4/1") | _____no_output_____ | Apache-2.0 | notebooks/supervised/detection/object-detection-efficientdet-d4.ipynb | lucasdavid/algorithms-in-tensorflow |
Application | INPUT_SHAPE = [299, 299, 3]
DATA_DIR = 'images/'
IMAGES = [
'https://raw.githubusercontent.com/keisen/tf-keras-vis/master/examples/images/goldfish.jpg',
'https://raw.githubusercontent.com/keisen/tf-keras-vis/master/examples/images/bear.jpg',
'https://raw.githubusercontent.com/keisen/tf-keras-vis/master/examples/images/soldiers.jpg',
'https://3.bp.blogspot.com/-W__wiaHUjwI/Vt3Grd8df0I/AAAAAAAAA78/7xqUNj8ujtY/s400/image02.png'
]
#@title
os.makedirs(os.path.join(DATA_DIR, 'unknown'), exist_ok=True)
for i in IMAGES:
_, f = os.path.split(i)
download_image(i, os.path.join(DATA_DIR, 'unknown', f))
images_set = (
tf.keras.preprocessing.image_dataset_from_directory(
DATA_DIR,
image_size=INPUT_SHAPE[:2],
batch_size=32,
shuffle=False)
.cache()
.prefetch(buffer_size=tf.data.experimental.AUTOTUNE))
#@title
plt.figure(figsize=(12, 4))
for images, _ in images_set.take(1):
for i, image in enumerate(images):
plt.subplot(math.ceil(len(images) / 4), 4, i+1)
plt.imshow(image.numpy().astype('uint8'))
plt.axis('off')
plt.tight_layout()
inputs = tf.cast(images[3:4], tf.uint8)
y = detector(inputs)
class_ids = y["detection_classes"]
print(*y.keys(), sep='\n')
y['num_detections']
y['detection_classes']
y['detection_boxes']
import matplotlib.patches as patches
fig, ax = plt.subplots(1)
ax.imshow(inputs[0].numpy())
# for y['detection_boxes']:
ax.add_patch(patches.Rectangle((50,100),40,30,linewidth=1,edgecolor='r',facecolor='none'))
; | _____no_output_____ | Apache-2.0 | notebooks/supervised/detection/object-detection-efficientdet-d4.ipynb | lucasdavid/algorithms-in-tensorflow |
Imports | from IPython.display import clear_output
!pip install path.py
!pip install pytorch3d
clear_output()
import numpy as np
import math
import random
import os
import plotly.graph_objects as go
import plotly.express as px
import torch
from torch.utils.data import Dataset, DataLoader, Subset
from torchvision import transforms, utils
from path import Path
random.seed = 42
!wget http://3dvision.princeton.edu/projects/2014/3DShapeNets/ModelNet10.zip
!unzip -q ModelNet10.zip
path = Path("ModelNet10")
folders = [dir for dir in sorted(os.listdir(path)) if os.path.isdir(path/dir)]
clear_output()
classes = {folder: i for i, folder in enumerate(folders)}
classes
def default_transforms():
return transforms.Compose([
PointSampler(1024),
Normalize(),
RandomNoise(),
ToSorted(),
ToTensor()
])
!gdown https://drive.google.com/uc?id=1CVwVxdfUfP6TRcVUjjJvQeRcgCGcnSO_
from helping import *
clear_output() | _____no_output_____ | MIT | data_processing.ipynb | annwhoorma/pmldl-project |
Data Preprocessing (optional) | with open(path/"dresser/train/dresser_0001.off", 'r') as f:
verts, faces = read_off(f)
i, j, k = np.array(faces).T
x, y, z = np.array(verts).T
# len(x)
# visualize_rotate([go.Mesh3d(x=x, y=y, z=z, color='lightpink', opacity=0.50, i=i,j=j,k=k)]).show()
# visualize_rotate([go.Scatter3d(x=x, y=y, z=z, mode='markers')]).show()
# pcshow(x, y, z)
pointcloud = PointSampler(1024)((verts, faces))
# pcshow(*pointcloud.T)
norm_pointcloud = Normalize()(pointcloud)
# pcshow(*norm_pointcloud.T)
noisy_pointcloud = RandomNoise()(norm_pointcloud)
# pcshow(*noisy_pointcloud.T)
rot_pointcloud = RandomRotation_z()(noisy_pointcloud)
# pcshow(*rot_pointcloud.T)
sorted_pointcloud = ToSorted()(rot_pointcloud)
# pcshow(*sorted_pointcloud.T)
tensor_pointcloud = ToTensor()(sorted_pointcloud) | _____no_output_____ | MIT | data_processing.ipynb | annwhoorma/pmldl-project |
Creating Loaders for Final Progress Report Redefine classes | class PointCloudData(Dataset):
def __init__(self, root_dir, valid=False, folder="train", transform=default_transforms(), folders=None):
self.root_dir = root_dir
if not folders:
folders = [dir for dir in sorted(os.listdir(root_dir)) if os.path.isdir(root_dir/dir)]
self.classes = {folder: i for i, folder in enumerate(folders)}
self.transforms = transform
self.valid = valid
self.pcs = []
for category in self.classes.keys():
new_dir = root_dir/Path(category)/folder
for file in os.listdir(new_dir):
if file.endswith('.off'):
sample = {}
with open(new_dir/file, 'r') as f:
verts, faces = read_off(f)
sample['pc'] = (verts, faces)
sample['category'] = category
self.pcs.append(sample)
def __len__(self):
return len(self.pcs)
def __getitem__(self, idx):
pointcloud = self.transforms(self.pcs[idx]['pc'])
category = self.pcs[idx]['category']
return pointcloud, self.classes[category]
class PointCloudDataPre(Dataset):
def __init__(self, root_dir, valid=False, folder="train", transform=default_transforms(), folders=None):
self.root_dir = root_dir
if not folders:
folders = [dir for dir in sorted(os.listdir(root_dir)) if os.path.isdir(root_dir/dir)]
self.classes = {folder: i for i, folder in enumerate(folders)}
self.transforms = transform
self.valid = valid
self.pcs = []
for category in self.classes.keys():
new_dir = root_dir/Path(category)/folder
for file in os.listdir(new_dir):
if file.endswith('.off'):
sample = {}
with open(new_dir/file, 'r') as f:
verts, faces = read_off(f)
sample['pc'] = self.transforms((verts, faces))
sample['category'] = category
self.pcs.append(sample)
def __len__(self):
return len(self.pcs)
def __getitem__(self, idx):
pointcloud = self.pcs[idx]['pc']
category = self.pcs[idx]['category']
return pointcloud, self.classes[category]
class PointCloudDataBoth(Dataset):
def __init__(self, root_dir, valid=False, folder="train", static_transform=default_transforms(), later_transform=None, folders=None):
self.root_dir = root_dir
if not folders:
folders = [dir for dir in sorted(os.listdir(root_dir)) if os.path.isdir(root_dir/dir)]
self.classes = {folder: i for i, folder in enumerate(folders)}
self.static_transform = static_transform
self.later_transform = later_transform
self.valid = valid
self.pcs = []
for category in self.classes.keys():
new_dir = root_dir/Path(category)/folder
for file in os.listdir(new_dir):
if file.endswith('.off'):
sample = {}
with open(new_dir/file, 'r') as f:
verts, faces = read_off(f)
sample['pc'] = self.static_transform((verts, faces))
sample['category'] = category
self.pcs.append(sample)
def __len__(self):
return len(self.pcs)
def __getitem__(self, idx):
pointcloud = self.pcs[idx]['pc']
if self.later_transform is not None:
pointcloud = self.later_transform(pointcloud)
category = self.pcs[idx]['category']
return pointcloud, self.classes[category]
!mkdir drive/MyDrive/Thesis/dataloaders/final | _____no_output_____ | MIT | data_processing.ipynb | annwhoorma/pmldl-project |
Overfitting - all augmentations applied before training | BATCH_SIZE = 48
trs = transforms.Compose([
PointSampler(1024),
ToSorted(),
Normalize(),
ToTensor()
])
beds_train_dataset = PointCloudDataPre(path, folders=['bed'], transform=trs)
beds_valid_dataset = PointCloudDataPre(path, folder='test', folders=['bed'], transform=trs)
beds_train_loader = DataLoader(dataset=beds_train_dataset, shuffle=True, batch_size=BATCH_SIZE, drop_last=True)
beds_valid_loader = DataLoader(dataset=beds_valid_dataset, batch_size=BATCH_SIZE, drop_last=True)
!mkdir dataloader_beds_pre
torch.save(beds_train_loader, 'dataloader_beds_pre/trainloader.pth')
torch.save(beds_valid_loader, 'dataloader_beds_pre/validloader.pth')
!mkdir drive/MyDrive/Thesis/dataloaders/final
!cp -r dataloader_beds_pre drive/MyDrive/Thesis/dataloaders/final | mkdir: cannot create directory ‘dataloader_beds_pre’: File exists
mkdir: cannot create directory ‘drive/MyDrive/Thesis/dataloaders/final’: File exists
| MIT | data_processing.ipynb | annwhoorma/pmldl-project |
Underfitting - all augmentations applied during training | BATCH_SIZE = 48
trs = transforms.Compose([
PointSampler(1024),
ToSorted(),
Normalize(),
RandomNoise(),
ToTensor()
])
beds_train_dataset = PointCloudData(path, folders=['bed'], transform=trs)
beds_valid_dataset = PointCloudData(path, folder='test', folders=['bed'], transform=trs)
beds_train_loader = DataLoader(dataset=beds_train_dataset, num_workers=4, shuffle=True, batch_size=BATCH_SIZE, drop_last=True)
beds_valid_loader = DataLoader(dataset=beds_valid_dataset, num_workers=4, batch_size=BATCH_SIZE, drop_last=True)
!mkdir dataloader_beds_dur
torch.save(beds_train_loader, 'dataloader_beds_dur/trainloader.pth')
torch.save(beds_valid_loader, 'dataloader_beds_dur/validloader.pth')
!cp -r dataloader_beds_dur drive/MyDrive/Thesis/dataloaders/final | /usr/local/lib/python3.7/dist-packages/torch/utils/data/dataloader.py:481: UserWarning:
This DataLoader will create 4 worker processes in total. Our suggested max number of worker in current system is 2, which is smaller than what this DataLoader is going to create. Please be aware that excessive worker creation might get DataLoader running slow or even freeze, lower the worker number to avoid potential slowness/freeze if necessary.
| MIT | data_processing.ipynb | annwhoorma/pmldl-project |
Both - static and dynamic transformations | BATCH_SIZE = 48
static_trs = transforms.Compose([
PointSampler(1024),
ToSorted(),
Normalize(),
])
dynamic_trs = transforms.Compose([
RandomNoise(),
ToTensor()
])
beds_train_dataset = PointCloudDataBoth(path, folders=['bed'], static_transform=static_trs, later_transform=dynamic_trs)
beds_valid_dataset = PointCloudDataBoth(path, folder='test', folders=['bed'], static_transform=static_trs)
beds_train_loader = DataLoader(dataset=beds_train_dataset, shuffle=True, batch_size=BATCH_SIZE, drop_last=True)
beds_valid_loader = DataLoader(dataset=beds_valid_dataset, batch_size=BATCH_SIZE, drop_last=True)
!mkdir dataloader_beds_both
torch.save(beds_train_loader, 'dataloader_beds_both/trainloader.pth')
torch.save(beds_valid_loader, 'dataloader_beds_both/validloader.pth')
!cp -r dataloader_beds_both drive/MyDrive/Thesis/dataloaders/final | mkdir: cannot create directory ‘dataloader_beds_both’: File exists
| MIT | data_processing.ipynb | annwhoorma/pmldl-project |
Two classes: beds and tables | BATCH_SIZE = 48
static_trs = transforms.Compose([
PointSampler(1024),
ToSorted(),
Normalize(),
])
dynamic_trs = transforms.Compose([
RandomNoise(),
ToTensor()
])
beds_train_dataset = PointCloudDataBoth(path, folders=['bed', 'table'], static_transform=static_trs, later_transform=dynamic_trs)
beds_valid_dataset = PointCloudDataBoth(path, folder='test', folders=['bed', 'table'], static_transform=trs)
beds_train_loader = DataLoader(dataset=beds_train_dataset, shuffle=True, batch_size=BATCH_SIZE, drop_last=True)
beds_valid_loader = DataLoader(dataset=beds_valid_dataset, batch_size=BATCH_SIZE, drop_last=True)
!mkdir dataloader_beds_tables
torch.save(beds_train_loader, 'dataloader_beds_tables/trainloader.pth')
torch.save(beds_valid_loader, 'dataloader_beds_tables/validloader.pth')
!cp -r dataloader_beds_tables drive/MyDrive/Thesis/dataloaders/final | _____no_output_____ | MIT | data_processing.ipynb | annwhoorma/pmldl-project |
For 512 | !mkdir drive/MyDrive/Thesis/dataloaders/final512 | _____no_output_____ | MIT | data_processing.ipynb | annwhoorma/pmldl-project |
Overfitting - all augmentations applied before training | BATCH_SIZE = 48
trs = transforms.Compose([
PointSampler(512),
ToSorted(),
Normalize(),
ToTensor()
])
beds_train_dataset = PointCloudDataPre(path, folders=['bed'], transform=trs)
beds_valid_dataset = PointCloudDataPre(path, folder='test', folders=['bed'], transform=trs)
beds_train_loader = DataLoader(dataset=beds_train_dataset, shuffle=True, batch_size=BATCH_SIZE, drop_last=True)
beds_valid_loader = DataLoader(dataset=beds_valid_dataset, batch_size=BATCH_SIZE, drop_last=True)
!mkdir dataloader_beds_pre
torch.save(beds_train_loader, 'dataloader_beds_pre/trainloader.pth')
torch.save(beds_valid_loader, 'dataloader_beds_pre/validloader.pth')
!mkdir drive/MyDrive/Thesis/dataloaders/final
!cp -r dataloader_beds_pre drive/MyDrive/Thesis/dataloaders/final512 | mkdir: cannot create directory ‘dataloader_beds_pre’: File exists
mkdir: cannot create directory ‘drive/MyDrive/Thesis/dataloaders/final’: File exists
| MIT | data_processing.ipynb | annwhoorma/pmldl-project |
Underfitting - all augmentations applied during training | BATCH_SIZE = 48
trs = transforms.Compose([
PointSampler(512),
ToSorted(),
Normalize(),
ToTensor()
])
beds_train_dataset = PointCloudData(path, folders=['bed'], transform=trs)
beds_valid_dataset = PointCloudData(path, folder='test', folders=['bed'], transform=trs)
beds_train_loader = DataLoader(dataset=beds_train_dataset, num_workers=4, shuffle=True, batch_size=BATCH_SIZE, drop_last=True)
beds_valid_loader = DataLoader(dataset=beds_valid_dataset, num_workers=4, batch_size=BATCH_SIZE, drop_last=True)
!mkdir dataloader_beds_dur
torch.save(beds_train_loader, 'dataloader_beds_dur/trainloader.pth')
torch.save(beds_valid_loader, 'dataloader_beds_dur/validloader.pth')
!cp -r dataloader_beds_dur drive/MyDrive/Thesis/dataloaders/final512 | /usr/local/lib/python3.7/dist-packages/torch/utils/data/dataloader.py:481: UserWarning:
This DataLoader will create 4 worker processes in total. Our suggested max number of worker in current system is 2, which is smaller than what this DataLoader is going to create. Please be aware that excessive worker creation might get DataLoader running slow or even freeze, lower the worker number to avoid potential slowness/freeze if necessary.
| MIT | data_processing.ipynb | annwhoorma/pmldl-project |
Both - static and dynamic transformations | BATCH_SIZE = 48
static_trs = transforms.Compose([
PointSampler(512),
ToSorted(),
Normalize(),
])
dynamic_trs = transforms.Compose([
RandomNoise(),
ToTensor()
])
beds_train_dataset = PointCloudDataBoth(path, folders=['bed'], static_transform=static_trs, later_transform=dynamic_trs)
beds_valid_dataset = PointCloudDataBoth(path, folder='test', folders=['bed'], static_transform=static_trs)
beds_train_loader = DataLoader(dataset=beds_train_dataset, shuffle=True, batch_size=BATCH_SIZE, drop_last=True)
beds_valid_loader = DataLoader(dataset=beds_valid_dataset, batch_size=BATCH_SIZE, drop_last=True)
!mkdir dataloader_beds_both
torch.save(beds_train_loader, 'dataloader_beds_both/trainloader.pth')
torch.save(beds_valid_loader, 'dataloader_beds_both/validloader.pth')
!cp -r dataloader_beds_both drive/MyDrive/Thesis/dataloaders/final512 | mkdir: cannot create directory ‘dataloader_beds_both’: File exists
| MIT | data_processing.ipynb | annwhoorma/pmldl-project |
Two classes: beds and tables | BATCH_SIZE = 48
static_trs = transforms.Compose([
PointSampler(512),
ToSorted(),
Normalize(),
])
dynamic_trs = transforms.Compose([
RandomNoise(),
ToTensor()
])
beds_train_dataset = PointCloudDataBoth(path, folders=['bed', 'table'], static_transform=static_trs, later_transform=dynamic_trs)
beds_valid_dataset = PointCloudDataBoth(path, folder='test', folders=['bed', 'table'], static_transform=trs)
beds_train_loader = DataLoader(dataset=beds_train_dataset, shuffle=True, batch_size=BATCH_SIZE, drop_last=True)
beds_valid_loader = DataLoader(dataset=beds_valid_dataset, batch_size=BATCH_SIZE, drop_last=True)
!mkdir dataloader_beds_tables
torch.save(beds_train_loader, 'dataloader_beds_tables/trainloader.pth')
torch.save(beds_valid_loader, 'dataloader_beds_tables/validloader.pth')
!cp -r dataloader_beds_tables drive/MyDrive/Thesis/dataloaders/final | _____no_output_____ | MIT | data_processing.ipynb | annwhoorma/pmldl-project |
Fill in full author references for mentions of authorsFor example, if we find `Calderon`, we want to produce a string```Pedro Calderón de la Barca``` | testTexts = [
"Calderón de la Barca, Pedro",
"CCCCCalderón",
"Caldeeeeeerón",
"Pedro Barca",
"Pedro Barca",
"Agustin Moreto",
"A. Moreto",
"Agustin",
"Augustine",
] | _____no_output_____ | MIT | pyling/detectAuthors.ipynb | dirkroorda/explore |
TriggersWe are going to find trigger strings for authors in the input texts.In order to do that successfully, we normalize the text first:* we remove all accents from accented letters* we make everything lowercase We need a function that can strip accents from characters.From [stackoverflow](https://stackoverflow.com/questions/517923/what-is-the-best-way-to-remove-accents-in-a-python-unicode-string) | import re
import unicodedata
def normalize(text):
text = unicodedata.normalize('NFD', text)
text = text.encode('ascii', 'ignore')
text = text.decode("utf-8")
return text.lower().strip()
normalize("Calderón de la Barca, Pedro") | _____no_output_____ | MIT | pyling/detectAuthors.ipynb | dirkroorda/explore |
AuthorsWe compile a list of authors that we want to detect.For each author we have a full name, a key, and a list of triggers.We format the specficiation as a *yaml* file (which maps to a Python dictionary). | authorSpec = '''
cald:
full: Pedro Calderón de la Barca
triggers:
- calderon
- barca
more:
full: Agustín Moreto
triggers:
- moreto
- agustin
- augustine
''' | _____no_output_____ | MIT | pyling/detectAuthors.ipynb | dirkroorda/explore |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.