NeurIPS 2020

Collegial Ensembles

Meta Review

This paper explores ensembles from the perspective of neural networks in the width limit. The reviewers all found that the paper is well written, technically sound and the contributions are novel and significant. There was significant discussion following the author rebuttal and multiple reviewers were willing to champion the paper for acceptance. One concern shared by reviewers was the motivation for why "var(K)" was a quantity that that should be minimized, as was presented in the theory of the paper. Overall, this seems like an exciting paper that will be of interest at the conference.