SHAP, LIME, feature importance, and model interpretability
Permutation importance, partial dependence plots (PDP), individual conditional expectation (ICE), feature interaction, and built-in vs model-agnostic importance
SHAP theory (Shapley values from game theory), TreeSHAP, KernelSHAP, DeepSHAP, LIME (local linear approximations), anchors, and when to use each
Error analysis, slice-based evaluation, adversarial examples, robustness testing, model calibration (Platt scaling, isotonic), and uncertainty estimation (MC Dropout, ensembles)
Types of bias (selection, measurement, algorithmic), fairness metrics (demographic parity, equalized odds, calibration), bias detection (AIF360, Fairlearn), and mitigation strategies (pre/in/post-processing)