Identifying the Most Explainable Classifier

Brett Mullins

We introduce the notion of pointwise coverage to measure the explainability properties of machine learning classifiers. An explanation for a prediction is a definably simple region…