1 · Linear regression
Fit a line to data — then a polynomial — then regularize. The simplest model with the deepest lessons.
●Track 5 · live
Classical Supervised ML
Linear regression to XGBoost — the workhorses you'll reach for first.
Linear and polynomial regression, logistic regression, SVM, Naive Bayes, KNN, decision trees, random forests, AdaBoost, Gradient Boosting, XGBoost. Each algorithm comes with a cinematic scene (loss surface, decision boundary, ensemble vote) plus a sklearn-in-browser implementation.
28 / 50 lessons8 chapters
2 · Metrics & validation
Pick the right error metric, validate honestly with k-fold CV, and diagnose bias vs variance — three skills that separate amateur from professional ML.
3 · Logistic regression & classification
Sigmoid + BCE for binary, softmax for multiclass, plus the metrics (precision, recall, F1, ROC-AUC) and tuning that turn a fit into a deployable model.
4 · Support Vector Machines
Hard + soft margins, the kernel trick, and SVR — the geometric classifier that ruled tabular ML before deep learning.
5 · Naive Bayes & KNN
Two classic non-linear classifiers — the speed-king Naive Bayes and the lazy-but-powerful K-Nearest Neighbors.
6 · Decision trees
Recursive splits by impurity, regression by variance reduction, and pruning to fight overfitting — the atomic unit of every ensemble.
7 · Ensemble methods
Bagging, random forest, and AdaBoost — the algorithms that turn weak learners into strong learners.
8 · Gradient boosting & capstone
Gradient boosting, XGBoost, and a full end-to-end ML pipeline — the deliverable shape of any production model.