异构编码联邦学习  

Heterogeneous coding-based federated learning

作　　者：史洪玮[1] 洪道诚 施连敏[3,4] 杨迎尧 SHI Hongwei;HONG Daocheng;SHI Lianmin;YANG Yingyao(School of Information Engineering,Suqian University,Suqian,Jiangsu 223800,China;Shanghai Institute of AI for Education&School of Computer Science and Technology,East China Normal University,Shanghai 200062,China;School of Computer Science and Technology,Soochow University,Suzhou,Jiangsu 215008,China;The Key Laboratory of Cognitive Computing and Intelligent Information Processing of Fujian Education Institutions,Wuyi University,Wuyishan,Fujian 354300,China)

机构地区：[1]宿迁学院信息工程学院,江苏宿迁223800 [2]华东师范大学上海智能教育研究院&计算机科学与技术学院,上海200062 [3]苏州大学计算机科学与技术学院,江苏苏州215008 [4]武夷学院认知计算与智能信息处理福建省高校重点实验室,福建武夷山354300

出　　处：《华东师范大学学报（自然科学版）》2023年第5期110-121,共12页Journal of East China Normal University(Natural Science)

基　　金：国家自然科学基金(61977025);2021年度江苏省重点研发计划(现代农业)项目(BE2021354);2020宿迁市项目(Z2020133);2021宿迁市现代农业项目(L202109);福建省高校重点实验室开放课题基金(KLCCIIP2021201);苏州市科技计划项目(SNG201908)。

摘　　要：异构联邦学习系统中的个人电脑、嵌入式设备等多种边缘设备,存在资源受限的掉队者设备降低联邦学习系统训练效率的问题.针对此问题,本文提出了异构编码联邦学习(heterogeneous coded-based federated learning,HCFL)系统框架,以实现:(1)提高系统训练效率,加快多掉队者场景下的异构联邦学习(federated learning,FL)训练速度;(2)提供一定级别的数据隐私保护.HCFL方案分别从客户端和服务器角度出发设计了调度策略,以满足通用环境下多掉队者模型计算加速;同时设计了线性编码计算方案(linear coded computing,LCC)为任务分发提供数据保护.实验结果表明,当异构FL中设备之间性能差异较大时,HCFL能够将训练时间缩短89.85%.In heterogeneous federated learning systems,among a variety of edge devices such as personal computers and embedded devices,resource-constrained devices,i.e.stragglers,reduce the training efficiency of the federated learning system.This paper proposes a heterogeneous coded federated learning(HCFL)system to①improve the training efficiency of the system and speed up the training of heterogeneous federated learning(FL)for multiple stragglers,②provide a certain level of data privacy protection.The HCFL scheme designs scheduling strategies from the perspective of client and server to satisfy the accelerated calculation of multiple stragglers model in the general environment.In addition,a linear coded computing(LCC)scheme is designed to provide data protection for task distribution.The experimental results show that HCFL can reduce training time by 89.85%when the performance difference between devices is large.

关 键 词：联邦学习 线性编码 异构系统 调度算法 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]