Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
How to use scikit-learn to solve machine learni...
Search
Olivier Grisel
April 22, 2015
Technology
0
990
How to use scikit-learn to solve machine learning problems
AutoML Hackathon - Paris - April 2015
Olivier Grisel
April 22, 2015
Tweet
Share
More Decks by Olivier Grisel
See All by Olivier Grisel
Intro to scikit-learn
ogrisel
5
670
An Intro to Deep Learning
ogrisel
1
240
Predictive Modeling and Deep Learning
ogrisel
2
340
Intro to scikit-learn and what's new in 0.17
ogrisel
1
330
Big Data, Predictive Modeling and tools
ogrisel
2
250
Recent Developments in Deep Learning
ogrisel
3
660
Documentation
ogrisel
2
210
Build and test wheel packages on Linux, OSX and Windows
ogrisel
2
330
Big Data and Predictive Modeling
ogrisel
3
220
Other Decks in Technology
See All in Technology
プルリクエストレビューを終わらせるためのチーム体制 / The Team for Completing Pull Request Reviews
nekonenene
3
1.5k
AI自体のOps 〜LLMアプリの運用、AWSサービスとOSSの使い分け〜
minorun365
PRO
9
1.1k
Ruby on Railsで持続可能な開発を行うために取り組んでいること
am1157154
3
170
4th place solution Eedi - Mining Misconceptions in Mathematics
rist
0
150
"TEAM"を導入したら最高のエンジニア"Team"を実現できた / Deploying "TEAM" and Building the Best Engineering "Team"
yuj1osm
1
240
目標と時間軸 〜ベイビーステップでケイパビリティを高めよう〜
kakehashi
PRO
8
1k
エンジニアのキャリアパスと、 その中で自分が大切にしていること
noteinc
3
230
Cracking the Coding Interview 6th Edition
gdplabs
14
28k
遷移の高速化 ヤフートップの試行錯誤
narirou
6
1.9k
MIMEと文字コードの闇
hirachan
2
1.5k
事業を差別化する技術を生み出す技術
pyama86
2
540
Two Blades, One Journey: Engineering While Managing
ohbarye
4
2.7k
Featured
See All Featured
Statistics for Hackers
jakevdp
797
220k
Documentation Writing (for coders)
carmenintech
68
4.6k
[RailsConf 2023 Opening Keynote] The Magic of Rails
eileencodes
28
9.3k
A Tale of Four Properties
chriscoyier
158
23k
Reflections from 52 weeks, 52 projects
jeffersonlam
348
20k
The Cost Of JavaScript in 2023
addyosmani
47
7.4k
Typedesign – Prime Four
hannesfritz
41
2.5k
Facilitating Awesome Meetings
lara
53
6.3k
Building a Scalable Design System with Sketch
lauravandoore
461
33k
What’s in a name? Adding method to the madness
productmarketing
PRO
22
3.3k
Rebuilding a faster, lazier Slack
samanthasiow
80
8.9k
Code Reviewing Like a Champion
maltzj
521
39k
Transcript
How to use scikit-learn to solve machine learning problems AutoML
Hackathon April 2015
Outline • Machine Learning refresher • scikit-learn • Demo: interactive
predictive modeling on Census Data with IPython notebook / pandas / scikit-learn • Combining models with Pipeline and parameter search
Predictive modeling ~= machine learning • Make predictions of outcome
on new data • Extract the structure of historical data • Statistical tools to summarize the training data into a executable predictive model • Alternative to hard-coded rules written by experts
type (category) # rooms (int) surface (float m2) public trans
(boolean) Apartment 3 50 TRUE House 5 254 FALSE Duplex 4 68 TRUE Apartment 2 32 TRUE
type (category) # rooms (int) surface (float m2) public trans
(boolean) Apartment 3 50 TRUE House 5 254 FALSE Duplex 4 68 TRUE Apartment 2 32 TRUE sold (float k€) 450 430 712 234
type (category) # rooms (int) surface (float m2) public trans
(boolean) Apartment 3 50 TRUE House 5 254 FALSE Duplex 4 68 TRUE Apartment 2 32 TRUE sold (float k€) 450 430 712 234 features target samples (train)
type (category) # rooms (int) surface (float m2) public trans
(boolean) Apartment 3 50 TRUE House 5 254 FALSE Duplex 4 68 TRUE Apartment 2 32 TRUE sold (float k€) 450 430 712 234 features target samples (train) Apartment 2 33 TRUE House 4 210 TRUE samples (test) ? ?
Training text docs images sounds transactions Labels Machine Learning Algorithm
Model Predictive Modeling Data Flow Feature vectors
New text doc image sound transaction Model Expected Label Predictive
Modeling Data Flow Feature vector Training text docs images sounds transactions Labels Machine Learning Algorithm Feature vectors
Inventory forecasting & trends detection Predictive modeling in the wild
Personalized radios Fraud detection Virality and readers engagement Predictive maintenance Personality matching
• Library of Machine Learning algorithms • Focus on established
methods (e.g. ESL-II) • Open Source (BSD) • Simple fit / predict / transform API • Python / NumPy / SciPy / Cython • Model Assessment, Selection & Ensembles
Train data Train labels Model Fitted model Test data Predicted
labels Test labels Evaluation model = ModelClass(**hyperparams) model.fit(X_train, y_train)
Train data Train labels Model Fitted model Test data Predicted
labels Test labels Evaluation model = ModelClass(**hyperparams) model.fit(X_train, y_train) y_pred = model.predict(X_test)
Train data Train labels Model Fitted model Test data Predicted
labels Test labels Evaluation model = ModelClass(**hyperparams) model.fit(X_train, y_train) y_pred = model.predict(X_test) accuracy_score(y_test, y_pred)
Support Vector Machine from sklearn.svm import SVC model = SVC(kernel="rbf",
C=1.0, gamma=1e-4) model.fit(X_train, y_train) y_predicted = model.predict(X_test) from sklearn.metrics import f1_score f1_score(y_test, y_predicted)
Linear Classifier from sklearn.linear_model import SGDClassifier model = SGDClassifier(alpha=1e-4, penalty="elasticnet")
model.fit(X_train, y_train) y_predicted = model.predict(X_test) from sklearn.metrics import f1_score f1_score(y_test, y_predicted)
Random Forests from sklearn.ensemble import RandomForestClassifier model = RandomForestClassifier(n_estimators=200) model.fit(X_train,
y_train) y_predicted = model.predict(X_test) from sklearn.metrics import f1_score f1_score(y_test, y_predicted)
None
None
Demo time! http://nbviewer.ipython.org/github/ogrisel/notebooks/blob/ master/sklearn_demos/Income%20classification.ipynb https://github.com/ogrisel/notebooks
Combining Models from sklearn.preprocessing import StandardScaler from sklearn.decomposition import RandomizedPCA
from sklearn.svm import SVC scaler = StandardScaler() X_train_scaled = scaler.fit_transform(X_train) pca = RandomizedPCA(n_components=10) X_train_pca = pca.fit_transform(X_train_scaled) svm = SVC(C=0.1, gamma=1e-3) svm.fit(X_train_pca, y_train)
Pipeline from sklearn.preprocessing import StandardScaler from sklearn.decomposition import RandomizedPCA from
sklearn.svm import SVC from sklearn.pipeline import make_pipeline pipeline = make_pipeline( StandardScaler(), RandomizedPCA(n_components=10), SVC(C=0.1, gamma=1e-3), ) pipeline.fit(X_train, y_train)
Scoring manually stacked models scaler = StandardScaler() X_train_scaled = scaler.fit_transform(X_train)
pca = RandomizedPCA(n_components=10) X_train_pca = pca.fit_transform(X_train_scaled) svm = SVC(C=0.1, gamma=1e-3) svm.fit(X_train_pca, y_train) X_test_scaled = scaler.transform(X_test) X_test_pca = pca.transform(X_test_scaled) y_pred = svm.predict(X_test_pca) accuracy_score(y_test, y_pred)
Scoring a pipeline pipeline = make_pipeline( RandomizedPCA(n_components=10), SVC(C=0.1, gamma=1e-3), )
pipeline.fit(X_train, y_train) y_pred = pipeline.predict(X_test) accuracy_score(y_test, y_pred)
Parameter search import numpy as np from sklearn.grid_search import RandomizedSearchCV
params = { 'randomizedpca__n_components': [5, 10, 20], 'svc__C': np.logspace(-3, 3, 7), 'svc__gamma': np.logspace(-6, 0, 7), } search = RandomizedSearchCV(pipeline, params, n_iter=30, cv=5) search.fit(X_train, y_train) # search.best_params_, search.grid_scores_
Thank you! • http://scikit-learn.org • https://github.com/scikit-learn/scikit-learn @ogrisel