联邦学习任务在片上网络系统中基于DVFS的可靠映射机制分析  

Analysis of Reliable Mapping Mechanism Based on DVFS for Federated Learning Tasks in On Chip Network Systems

作　　者：周宏[1] 吴霁霖 高炜 张靖怡 ZHOU Hong;WU Jilin;GAO Wei;ZHANG Jingyi(China National Tobacco Corporation Guizhou Provincial Company Information Center,Guizhou 550001,China;School of Automation,Southeast University,Jiangsu 215009,China)

机构地区：[1]中国烟草总公司贵州省公司信息中心,贵州550001 [2]东南大学自动化学院,江苏215009

出　　处：《集成电路应用》2024年第7期86-87,共2页Application of IC

基　　金：中国烟草总公司贵州省公司科技项目(2022XM27)。

摘　　要：阐述在片上网络平台上联邦学习任务的分配与调度机制。通过将联邦学习建模为任务DAG图,并综合考虑平台计算资源、通信能力,实现在实时性、相关性、执行可靠性约束下的任务优化映射。This paper describes the allocation and scheduling mechanism of federated learning tasks on on-chip network platforms.It models federated learning as a task DAG graph and comprehensively considers platform computing resources and communication capabilities to achieve task optimization mapping under real-time,correlation,and execution reliability constraints.

关 键 词：联邦学习 任务映射 片上网络 DVFS 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]