Navigating ML Explainability: From Theory to Implementation
summary
Overview of practical explainability: regulatory hooks are the EU AI Act and GDPR, and explainability must be designed from project start through feature engineering, not just after training. Concrete tools covered: SHAP and LIME, partial dependence and ALE plots, variance‑based sensitivity and feature interaction checks, plus model examples (logistic regression, random forest, XGBoost, neural nets). Warnings: PDP breaks with correlated features and LIME shows local instability; choose simpler models when accuracy gains are marginal.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 20:15Why explainability matters
- 25:34Regulatory landscape overview
- 35:38Explainability requirements
- 45:47Interpretability vs explainability
- 1:02:28Global model behavior
- 1:17:34Local decision reasoning
- 1:19:44Final thoughts
- 1:28:06Q&A
related talks
24mWhy and How to Build Explainability into your ML Workflow
35mWhy Do You Need an Explainable Ai?
15mIntroduction to Explainable AI
37mExplaining model explainability
50mDehumanizing Agents: Why Explainability is Crucial in the LLM Era
16mBuilding Trust through Explainable AI
28mManaging AI Risk: Beyond Explainability
17mChallenges Towards Production-Ready Explainable Machine Learning