Machine Learning with Scikit-Learn
summary
Scikit-learn shows concrete recipes: tokenize with a CountVectorizer to build a 75,000-word vocabulary, vectorize text into sparse matrices, train a linear SVM and see training accuracy near perfect but test accuracy rise from ~84% to 88% after stronger regularization. Workflows use 5‑fold cross‑validation, grid and randomized search over hyperparameters, and Pipelines to avoid leakage. For big data use hashing trick and partial_fit streaming with SGD or minibatch learners; topic models (LDA), faster t‑SNE and new solvers (SAG) are available.
AI-written summary · about
chapters (13)
- 0:00Scikit-Learn tutorial intro
- 2:08Talk overview basics
- 6:44Unsupervised transformations
- 9:34Model complexity bias-variance
- 12:22Cross-validation evaluation
- 16:00Grid search tuning
- 22:13Sentiment analysis text
- 35:17ML pipelines building
- 44:18Randomized parameter search
- 53:57Efficient parameter search
- 1:00:21Out-of-core large datasets
- 1:07:34Scikit-Learn 0.17 features
- 1:14:27Scikit-Learn 0.18 roadmap
related talks
01h 24mMachine Learning with Text in scikit learn
35mScikit-Learn can do THAT?!
21mKeynote: Easier Machine Learning Thoughts From Scikit-Learn
42mMachine Learning Crash Course
30mKick-Start your Understanding of Machine Learning with Python
27mMachine Learning at Scale
57mHow To Evaluate and Improve Machine Learning Models
23mText mining online data with scikit-learn