I present an overview of interpretable machine learning. This includes interpretable models and model-agnostic explanations.
0:00 Introduction
0:20 Traditional machine learning
0:45 The big idea of interpretable/explainable machine learning: we want to know WHY?
3.08 Two types of explainable machine learning
5:15 Type 1 with examples: interpretable models
8:15 What are model-agnostic explanation methods/models?
9:40 Advantages of model-agnostic methods
11:00 Model-agnostic example: Partial dependence plot
13:40 Model-agnostic example: Permutation importances
16:00 Model-agnostic example: Global surrogate
18:00 Model-agnostic example: Local surrogate (LIME)
21:30 Model-agnostic example: Local model-agnostic model (SHAP)
23:00 Counterfactual explanations
24:45 Conclusions
26:00 Literature
