Differentially-Private Federated Linear Bandits

TL;DR

This paper introduces FedUCB, a differentially-private federated algorithm for cooperative contextual linear bandits, providing theoretical regret bounds and empirical performance in multi-agent settings.

Contribution

It presents FedUCB, a novel multi-agent private algorithm for federated linear bandits with rigorous privacy and regret guarantees.

Findings

Improved regret bounds for cooperative bandit learning

Strong differential privacy guarantees

Competitive empirical performance in multi-agent scenarios

Abstract

The rapid proliferation of decentralized learning systems mandates the need for differentially-private cooperative learning. In this paper, we study this in context of the contextual linear bandit: we consider a collection of agents cooperating to solve a common contextual bandit, while ensuring that their communication remains private. For this problem, we devise \textsc{FedUCB}, a multiagent private algorithm for both centralized and decentralized (peer-to-peer) federated learning. We provide a rigorous technical analysis of its utility in terms of regret, improving several results in cooperative bandit learning, and provide rigorous privacy guarantees as well. Our algorithms provide competitive performance both in terms of pseudoregret bounds and empirical benchmark performance in various multi-agent settings.