轻量级联邦学习平台LightFL的设计与实现  

Design and Implementation of Lightweight Federated Learning Platform LightFL

作　　者：刘若轩 高凌航 谢国云 张宇帆 李宏恩 武文媗 王灿 LIU Ruoxuan;GAO Linghang;XIE Guoyun;ZHANG Yufan;LI Hongen;WU Wenxuan;WANG Can(School of Applied Science,Beijing Information Science and Technology University,Beijing 100192,China)

机构地区：[1]北京信息科技大学理学院,北京100192

出　　处：《现代信息科技》2025年第6期71-74,共4页Modern Information Technology

基　　金：北京信息科技大学2024年大学生创新创业训练计划项目(S202411232607);北京信息科技大学“青年骨干教师”支持计划(YBT202450)。

摘　　要：现有的联邦学习平台通常较为复杂,需要用户进行深入学习才能掌握其功能。为解决这一问题,开发了一款轻量级联邦学习平台——LightFL,旨在降低联邦学习框架的使用门槛。该平台支持独立同分布和非独立同分布数据的模拟,并实现了经典联邦学习算法,如FedAvg、FedProx和MOON。通过模块化架构和轻量级代码设计,用户能够轻松开展联邦学习实验,且该平台便于部署和易于维护。与其他平台相比,该平台仅需用户编写少量代码就能配置联邦学习环境,简化了操作流程,降低了使用门槛,并为进一步开发新的联邦学习算法提供了便利。The existing Federated Learning platforms are generally complex and require in-depth learning to master their functions for users.To address this issue,a lightweight Federated Learning platform,LightFL,is developed with the aim of lowering the entry barrier of usage for Federated Learning frameworks.The platform supports simulations of both Independent and Identically Distributed and Non-Independent and Identically Distributed data,and implements classic Federated Learning algorithms such as FedAvg,FedProx,and MOON.Through a modular architecture and lightweight code design,users can easily engage in Federated Learning and experiments,and the platform is easy to deploy and maintain.Compared to other platforms,the platform can configure the Federated Learning environment with only a small amount of code to write for users.It simplifies the workflow,reduces the entry barrier of usage,and facilitates the further development of new Federated Learning algorithms.

关 键 词：轻量级平台 联邦学习 独立同分布 非独立同分布 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]