NeurIPS 2020

Election Coding for Distributed Learning: Protecting SignSGD against Byzantine Attacks

Meta Review

Overall, reviewers found this a solid contribution to the goal of designing communication efficient and robust distributed optimization methods. They continued to have questions about the significance and scope of the solution, but appreciated the authors' analysis and clarity, both in the paper and in the author response. We also thank the authors for the clear description of the NeurIPS 2019 reviews, which helped us appreciate the many problems solved in this research and that another set of reviews had already largely been addressed. Please read the reviews carefully and incorporate these comments into the final version as best you can.