Tier 1: Foundations

Tier này dạy gì?

Pipeline ML: data → feature → model → loss → train → evaluate.
Linear regression: closed-form (normal equation), gradient descent, ridge/lasso.
Logistic regression: sigmoid, cross-entropy; multi-class softmax; calibration + ROC/AUC.
Bias-variance trade-off, regularization (L1/L2, dropout, early stop), cross-validation.

Lesson 01

Data → feature → model → loss → train → evaluate; train/val/test split; bias-variance trade-off intro.

Lesson 02

y = Xw + b, MSE; closed-form normal equation vs gradient descent; ridge/lasso regularization; R².

Lesson 03

Sigmoid, binary classification, cross-entropy; multi-class softmax; calibration, ROC/AUC.

Lesson 04

Bias-variance trade-off; L1/L2, dropout, early stopping; k-fold CV, nested CV; learning curves.