Interpretable Machine Learning with Python Examples
summary
Covers practical, model‑agnostic explainers with Python demos: LIME fits local linear surrogates over a RandomForest on a Titanic demo (example instance explanations, ~80% model accuracy; LIME is fast-ish but unstable across samples), and SHAP uses XGBoost on the Adult income set (72% accuracy), delivering Shapley values, force and summary plots; SHAP has stronger game‑theory grounding and unified global+local views but is far more computationally expensive unless using tree explainer or GPUs.
AI-written summary · about
chapters (8)
- 0:00Introduction to talk
- 2:49What is interpretable ML
- 5:16Why interpretable ML is needed
- 10:13Relevance to different roles
- 17:59Types of interpretability
- 21:09LIME explained with demo
- 32:05SHAP explained with demo
- 42:22Comparison and conclusion
related talks
33mExplainable AI in the LIME-light
01h 36mOpen the Black Box: an Introduction to Model Interpretability with LIME and SHAP
27mInterpretability and explainability in machine learning
34mInterpreting Machine Learning Models: Why and How!
31mBeyond the Black Box: Interpreting ML Models with SHAP
16mBuilding Trust through Explainable AI
46mExplaining a Machine Learning blackbox
15mInterpretable Machine Learning with R