Achieving Linear Speedup in Asynchronous Federated Learning with Heterogeneous Clients

arXiv:2402.11198v1 Announce Type: new
Abstract: Federated learning (FL) is an emerging distributed training paradigm that aims to learn a common global model without exchanging or transferring the data that are stored locally at different clients. The Federated Averaging (FedAvg)-based algorithms have gained substantial popularity in FL to reduce the communication overhead, where each client conducts multiple localized iterations before communicating with a central server. In this paper, we focus on FL where the clients have diverse computation and/or communication capabilities. …

abstract algorithms arxiv asynchronous communication cs.dc cs.lg data distributed federated learning global learn linear paradigm reduce training type