Explainable AI (XAI)

tags: XAI

Machine Learning

• 00. Data Exploration
• 01. Intrinsically Interpretable Models
• 02. Local Interpretable Model-agnostic (LIME)
• 03. SHapley Additive exPlanations (SHAP) - Tree SHAP
• 04. SHapley Additive exPlanations (SHAP) - Kernel SHAP

Deep Learning

• 01. Perturbation-Based
  • Occlusion Sensitivity (Zeiler & Fergus 14)
• 02. Gradient-Based
  • Sensitivity Analysis (Simonyan et al. 14)
  • SmoothGrad (Smikov et al. 17)
  • Integrated Gradients (Sundararajan et al. 17)
• 03. Propagation-Based
  • LRP (Bach et al. 15)
• 04. CAM-Based
  • Grad-CAM (Ancona et al. 16)
  • Layer-CAM (Peng-Tao et al. 21)

Reference:

• deepfindr: xai-series
• keisen: tf-keras-vis

Kaggle

• Machine learning explainability