NeurIPS 2020

Online Non-Convex Optimization with Imperfect Feedback


Meta Review

This paper led to a lot of discussion between the reviewers and the AC. While the reviewers appreciated the technical contributions of the paper, the main concern was that computational requirements of the proposed method are not clear. A discussion about the computational requirements is also important while comparing to prior methods. Overall, presenting convincing examples where these algorithms can be instantiated and specifically discussing about the implementation of sampling oracles will make this paper significantly stronger.