基于随机客户端参与的无偏联邦学习激励机制设计（英文版）

英文标题：Incentive Mechanism Design for Unbiased Federated Learning with Randomized Client Participation中文摘要：本文提出了一种基于博弈论的激励机制，通过随机客户端参与度数，实现以不偏差的方式获得高性能模型，并分析了客户端的参与度数和异质性数据如何影响模型性能的收敛界。英文摘要：Incentive mechanism is crucial for federated learning (FL) when rationalclients do not have the same interests in the global model as the server.However, due to system heterogeneity and limited budget, it is generallyimpractical for the server to incentivize all clients to participate in alltraining rounds (kn