Review for NeurIPS paper: ICAM: Interpretable Classification via Disentangled Representations and Feature Attribution Mapping