Evaluating Prototype Explanations in Machine Learning
Prototype-based post-hoc explanations aim to make model predictions interpretable by presenting representative examples (prototypes) that illustrate how the model arrives at decisions. Their evaluation often relies on quantitative metrics such as fidelity (how closely prototypes approximate the model’s decision function), coverage (how much of the input space they represent), stability (whether explanations remain consistent under small perturbations), and diversity (ensuring prototypes capture … Read more