NeurIPS 2020

Boundary thickness and robustness in learning models

Meta Review

This proposal relies on the concept of boundary thickness to explain robustness and provides a convincing experimental study. The reviewing committee recommends the acceptance of this paper. We suggest the authors incorporate the additional points included in the rebuttal to the final version of the paper. In particular, given the proximity with the concept of margin and the vast literature on the topic, we recommend the authors clarify the differences as they did in the rebuttal.