classical-mlevaluationmethodology
Cross-validation and model selection
k-fold, stratified, time-series CV — honest estimates of generalisation and hyper-parameter search.
Уровни глубины
L0Intro~1ч
Explains train/validation/test split; reads a CV score.
L1Basics~6ч
Runs k-fold, stratified k-fold, leave-one-out; knows when stratification matters.
L2Working~10ч
Designs CV for grouped/time-series data; combines with nested CV for unbiased tuning.
L3Advanced~20ч
Analyses CV variance; applies bootstrap, adversarial validation; multiple-testing corrections.
L4Research~40ч
Theoretical CV guarantees, distribution-shift aware validation.