BriefGPT.xyz
Oct, 2020
差分隐私联合线性赌博机算法
Differentially-Private Federated Linear Bandits
HTML
PDF
Abhimanyu Dubey, Alex Pentland
TL;DR
本文就基于上下文线性赌博机的联邦学习问题提出了一种称为 FedUCB 的多代理私有算法，该算法在中央化和去中心化（点对点）联邦学习方案中均可使用，在保证通信隐私的同时，在后遗憾度和隐私保证方面表现出极强的实用性。
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
→