通信学报2023，Vol.44Issue(11)：79-93,15.DOI:10.11959/j.issn.1000-436x.2023196
异构边缘计算环境下异步联邦学习的节点分组与分时调度策略
Client grouping and time-sharing scheduling for asynchronous federated learning in heterogeneous edge computing environment
摘要
Abstract
To overcome the three key challenges of federated learning in heterogeneous edge computing,i.e.,edge heter-ogeneity,data Non-IID,and communication resource constraints,a grouping asynchronous federated learning(FedGA)mechanism was proposed.Edge nodes were divided into multiple groups,each of which performed global updated asyn-chronously with the global model,while edge nodes within a group communicate with the parameter server through time-sharing communication.Theoretical analysis established a quantitative relationship between the convergence bound of FedGA and the data distribution among the groups.A time-sharing scheduling magic mirror method(MMM)was pro-posed to optimize the completion time of a single round of model updating within a group.Based on both the theoretical analysis for FedGA and MMM,an effective grouping algorithm was designed for minimizing the overall training com-pletion time.Experimental results demonstrate that the proposed FedGA and MMM can reduce model training time by 30.1%～87.4%compared to the existing state-of-the-art methods.关键词
边缘计算/联邦学习/非独立同分布数据/异构性/收敛分析Key words
edge computing/federated learning/Non-IID/heterogeneity/convergence analysis分类
信息技术与安全科学引用本文复制引用
马千飘,贾庆民,刘建春,徐宏力,谢人超,黄韬..异构边缘计算环境下异步联邦学习的节点分组与分时调度策略[J].通信学报,2023,44(11):79-93,15.基金项目
国家自然科学基金资助项目(No.U1709217,No.61936015,No.92267301) Foundation Item:The National Natural Science Foundation of China(No.U1709217,No.61936015,No.92267301) （No.U1709217,No.61936015,No.92267301）
