← 🏠 Trang chính / AI-ML

Tier 2 — Classical Methods

Tree-based + kernel methods — vẫn dominate tabular data trong industry. KNN, decision tree, random forest, SVM, gradient boosting.

Tier này dạy gì?

KNN: instance-based, distance metrics, curse of dimensionality.
Decision tree: entropy, information gain, Gini, CART, pruning.
Random forest: bagging, OOB error, feature importance.
SVM: margin maximization, kernel trick (RBF, polynomial).
Gradient boosting: AdaBoost → GBM → XGBoost/LightGBM.

Lộ trình 4 bài

Lesson 01

KNN & Decision tree

k-Nearest Neighbors (Euclidean/Manhattan/cosine), curse of dim; Decision tree (entropy/Gini, info gain, CART), pruning.

▶ Visualization

Lesson 02

Random forest & bagging

Bagging, OOB error, feature importance; bias-variance interpretation; ExtraTrees.

▶ Visualization

Lesson 03

Margin maximization, support vectors; kernel trick (RBF, polynomial); soft margin C; one-vs-rest cho multi-class.

▶ Visualization

Lesson 04

Gradient boosting

AdaBoost intuition → Gradient Boosting framework; XGBoost, LightGBM, CatBoost briefly.

▶ Visualization