NeurIPS 2020

Understanding Double Descent Requires A Fine-Grained Bias-Variance Decomposition


Meta Review

This paper studies the double descent phenomenon in random features models. The authors disentangle the sources of variance contributing to the irregular bias-variance behavior of certain ML methods. All reviewers had a favorable assessment of the paper. The reviewers raised various technical concerns in their reviews including relationship with prior work but thought that the authors’ response adequately addressed these concerns and multiple reviewers raised their score. I concur with this assessment and recommend acceptance.