32. Model evaluation and selection

Choose metrics, design cross-validation, tune hyperparameters, analyze errors, test robustness, and compare models without misleading results.