- Block 11 — Classification algorithms
- Objective
- rigorously distinguish boosting from bagging, understand the sequential construction and error correction that underlies it, master AdaBoost and gradient boosting in their algorithmic formulation, work the weight update and the pseudo-residual update by hand, tune the learning rate, the tree count and the regularization knobs, apply early stopping without contaminating the test set, know the modern implementations XGBoost, LightGBM, CatBoost and
HistGradientBoostingClassifier including their native handling of categorical variables and missing values, extend the family to quantile regression and imbalanced problems, calibrate the probabilities it produces, and articulate the strengths and limitations of this family on tabular data. - Estimated duration
- 120 minutes
- Prerequisites
- chapters 030, 031, 032, 033, 034, 035, 037, and 038
- Associated quizzes
039.1-quiz-bagging-vs-boosting.md through 039.7-quiz-strengths-and-limits.md