Federated Recommendation System via Differential Privacy
Machine Learning
2020-05-19 v2 Information Theory
math.IT
Abstract
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.
Keywords
Cite
@article{arxiv.2005.06670,
title = {Federated Recommendation System via Differential Privacy},
author = {Tan Li and Linqi Song and Christina Fragouli},
journal= {arXiv preprint arXiv:2005.06670},
year = {2020}
}
Comments
This paper is accepted by 2020 IEEE International Symposium on Information Theory(ISIT 2020)