Snow White and the Dwarf Ensemble

A fairy tale introduction to the concept of Ensemble Learning

Transcript

1. Snow White and the Dwarf Ensemble
2. About Me Johannes Kolbe • Data Scientist at celebrate company ◦ Focus on Computer Vision ◦ Some Expertise in NLP • M.Sc. in Computer Science at TU Berlin • Hugging Face Fellow ◦ Leading CV Study Group on discord: https://huggingface.co/join/discord ◦ Past Study Groups: https://github.com/huggingface/community-events/ tree/main/computer-vision-study-group @johko@sigmoid.social linkedin.com/in/johko github.com/johko huggingface.co/johko
3. Once upon a time…
4. Where the Small Folk Lives
5. Tragedy and Remedy
6. Happily Ever After
7. A New Beginning
8. Old Friends
9. The Dwarfs Diggy Peppy Glimmer Gizmo
10. The Dwarfs Rustle Flint Picker
11. The Misery of the Dwarfs
12. Hard Voting Ruby Amethyst Ruby Garnet Amethyst Ruby Ruby 4x Ruby 2x Amethyst 1x Garnet P
13. Soft Voting 80% Ruby 15% Amethyst 5% Garnet 40% Amethyst 35% Ruby 25% Garnet 95% Ruby 3% Amethyst 2% Garnet 50% Garnet 40% Ruby 10% Amethyst 50% Amethyst 25% Garnet 25% Ruby 80% Ruby 15% Amethyst 5% Garnet 75% Ruby 15% Garnet 10% Amethyst 62% Ruby 20% Amethyst 18% Garnet P
14. Voting Classifier import numpy as np from sklearn.linear_model import LogisticRegression from sklearn.naive_bayes import GaussianNB from sklearn.ensemble import RandomForestClassifier, VotingClassifier clf1 = LogisticRegression(multi_class='multinomial', random_state=1) clf2 = RandomForestClassifier(n_estimators=50, random_state=1) clf3 = GaussianNB() X = np.array([[-1, -1], [-2, -1], [-3, -2], [1, 1], [2, 1], [3, 2]]) y = np.array([1, 1, 1, 2, 2, 2]) eclf1 = VotingClassifier(estimators=[('lr', clf1), ('rf', clf2), ('gnb', clf3)], voting='hard') eclf1 = eclf1.fit(X, y) eclf1.predict(X) https://scikit-learn.org/stable/modules/g enerated/sklearn.ensemble.VotingClas sifier.html
15. Bagging P
16. Bagging Classifier from sklearn.ensemble import BaggingClassifier from sklearn.datasets import make_classification X, y = make_classification(n_samples=100, n_features=4, n_informative=2, n_redundant=0, random_state=0, shuffle=False) clf = BaggingClassifier(n_estimators=10, random_state=0).fit(X, y) clf.predict([[0, 0, 0, 0]]) clf.score(X, y) https://scikit-learn.org/stable/modules/g enerated/sklearn.ensemble.RandomFo restClassifier.html#sklearn.ensemble.R andomForestClassifier
17. Boosting (AdaBoost) update weights update weights 80% 70% 55% 65% 90% 85% 80%
18. Boosting (AdaBoost) P
19. Adaboost Classifier from sklearn.ensemble import AdaBoostClassifier from sklearn.datasets import make_classification X, y = make_classification(n_samples=1000, n_features=4, n_informative=2, n_redundant=0, random_state=0, shuffle=False) clf = AdaBoostClassifier(n_estimators=100) clf.fit(X, y) clf.predict([[0, 0, 0, 0]]) clf.score(X, y) https://scikit-learn.org/stable/modules/g enerated/sklearn.ensemble.AdaBoostC lassifier.html#sklearn.ensemble.AdaBo ostClassifier
20. Stacking P P P Labels
21. Stacking Classifier from sklearn.datasets import load_iris from sklearn.ensemble import BaggingClassifier from sklearn.ensemble import AdaBoostClassifier from sklearn.linear_model import LogisticRegression from sklearn.ensemble import StackingClassifier X, y = load_iris(return_X_y=True) estimators = [ ('rf', BaggingClassifier(n_estimators=10, random_state=42)), ('svr', AdaBoostClassifier(n_estimators=100)) ] clf = StackingClassifier( estimators=estimators, final_estimator=LogisticRegression() ) from sklearn.model_selection import train_test_split X_train, X_test, y_train, y_test = train_test_split(X, y, stratify=y, random_state=42) clf.fit(X_train, y_train).score(X_test, y_test) https://scikit-learn.org/stable/modules/g enerated/sklearn.ensemble.StackingCl assifier.html
22. Thank You
23. Outtakes
24. Outtakes
25. Outtakes
26. CRÉDITOS: Esta plantilla para presentaciones es una creación de Slidesgo, e incluye iconos de Flaticon, infografías e imágenes de Freepik Gracia s ¿Tienes alguna pregunta? tuemail@freepik.com +91 620 421 838 tupaginaweb.com Por favor, conserva esta diapositiva para atribuirnos Slide Design from slidesgo: