Explainable Machine Learning

This is a field that attempts to provide oversight over AI algorithms. Explains how AI arrived to its specific decision.

Formal Definition

Given a Black Box model $f$, and explainable AI algorithm constructs in an Interpretable function $h$ which approximates $f$ as closely as possible.

Types

• Intrinsically Interpretable Models
• Model-Agnostic XAI
• Non-Model-Agnostic XAI

Techniques

• Saliency Map
• LIME
• SHAP
• Attention Map