文章目录

• 0 前言
• 1 Introduction of Explainable/ Interpretable ML
• • 1.1 Why we need Explainable ML?
  • 1.2 Interpretable v.s. Powerful
• 2 Local Explanation
• • 2.1 Introduction of Local Explanation
  • • 2.1.1 Removing Way
    • 2.1.2 Modifying Way
    • 2.1.3 Limitation of Gradient based Approaches
  • 2.2 Attack Interpretation
  • 2.3 Help by Saliency Map
• 3 Global Explanation
• • 3.1 Activation Minimization (review)
  • 3.2 "Regularization" from Generator
• 4 Using a model to explain another
• • 4.1 Using Linear Model
  • 4.2 Local Interpretable Model-Agnostic Explanations (LIME)
  • • 4.2.1 Basic Idea of LIME
    • 4.2.2 LIME-Image
  • 4.3 Using Decision Tree