classical-mlsupervisedtreesensembles
Gradient boosting
XGBoost, LightGBM, CatBoost — the workhorse of tabular ML competitions and production pipelines.
Уровни глубины
L0Intro~2ч
Reads a boosted-tree prediction; knows "many weak learners → strong model".
L1Basics~10ч
Derives AdaBoost; writes gradient-boosting pseudocode; tunes n_estimators, lr, depth.
L2Working~20ч
Ships XGBoost/LightGBM with CV, early stopping, monotonic constraints; handles categoricals with CatBoost.
L3Advanced~30ч
Reads histogram-based tree construction; implements custom loss & objective; SHAP interpretation.
L4Research~60ч
Contributes to tree-structure learning or boosting theory (margin bounds, functional gradient).
Ресурсы
L1 — Basics