NeurIPS 2020

Fast Adversarial Robustness Certification of Nearest Prototype Classifiers for Arbitrary Seminorms

Meta Review

Four experts in the field reviewed the paper. Two reviewers put the paper marginally above the acceptance threshold, and the other two placed it marginally below. The reviewers found the rebuttal useful, but they had lingering questions, such as the work's relationship to 1-NN. Considering the nature of the reviewers' comments, AC felt the concerns could be addressed in a paper revision. The decision is to recommend the paper for acceptance. AC suggested the authors addressing the lingering questions to their best ability when preparing the camera-ready.