Model Evaluation & Validation — Machine Learning Roadmap
Knowing whether a model will actually work on data it hasn't seen yet
Steps in Model Evaluation & Validation
- Train/Test/Validation Splits — advanced · Why data must be split before training, and what each split is for
- Cross-Validation Techniques — advanced · k-fold and stratified cross-validation for more reliable performance estimates
- Bias-Variance Tradeoff — advanced · Understanding the fundamental tension between underfitting and overfitting
- Overfitting & Underfitting — advanced · Recognizing the symptoms of each and the levers you can pull to fix them
- Hyperparameter Tuning (Grid Search, Random Search) — advanced · Systematically searching for the best model configuration
- Learning Curves & Diagnostics — advanced · Reading training/validation curves to decide what to fix next
Part of
- Machine Learning roadmap — the full learning path