|Federated Recommendation System via Differential Privacy
|Tan Li, Linqi Song, City University of Hong Kong, China; Christina Fragouli, University of California, Los Angeles, United States
|L.5: Federated Learning
|Statistics and Learning Theory
|Click here to download the manuscript
|Click here to watch in the Virtual Symposium
|In this paper we are interested in what we term the federated private bandits framework, that combines differential privacy with multi-agent bandit learning. We explore how differential privacy based Upper Confidence Bound (UCB) methods can be applied to multi-agent environments, and in particular to federated learning environments both in ‘master-worker’ and ‘fully decentralized’ settings. We provide a theoretical analysis on the privacy and regret performance of the proposed methods and explore the tradeoffs between these two.