NeurIPS 2020

Differentially-Private Federated Linear Bandits

Meta Review

Three of the four reviewers evaluated this paper very positively, noting the good, rigorous theory and the relevance of the topic of federated bandits with privacy guarantees, and judged the work as potentially high-impact. The author rebuttal was taken into account and the subsequent discussion led the 4th (less positive) reviewer to increase his/her score. Therefore, this paper is accepted. However, for the final version, the authors are asked to explicitly report the values of epsilon and delta in the experiments, as standard in the differential privacy literature.