Sound Explanation for Trustworthy Machine Learning

Kai Jia, Pasapol Saowakon, Limor Appelbaum +1

We take a formal approach to the explainability problem of machine learning systems. We argue against the practice of interpreting black-box models via attributing scores to input…