《计算机应用》唯一官方网站 ›› 2025, Vol. 45 ›› Issue (7): 2123-2131.DOI: 10.11772/j.issn.1001-9081.2024070934

• CCF第39届中国计算机应用大会 (CCF NCCA 2024) • 上一篇    下一篇

面向个性化与公平性的联邦学习算法

张宏扬1,2,3, 张淑芬1,2,4(), 谷铮1,2,3   

  1. 1.华北理工大学 理学院,河北 唐山 063210
    2.河北省数据科学与应用重点实验室(华北理工大学),河北 唐山 063210
    3.唐山市数据科学重点实验室(华北理工大学),河北 唐山 063210
    4.唐山市大数据安全与智能计算重点实验室(华北理工大学),河北 唐山 063210
  • 收稿日期:2024-07-05 修回日期:2024-09-25 接受日期:2024-09-29 发布日期:2025-07-10 出版日期:2025-07-10
  • 通讯作者: 张淑芬
  • 作者简介:张宏扬(1999—),男,江苏淮安人,硕士研究生,CCF会员,主要研究方向:数据安全、隐私保护
    张淑芬(1972—),女,河北唐山人,教授,硕士,CCF高级会员,主要研究方向:云计算、数据安全、隐私保护 zhsf@ncst.edu.cn
    谷铮(1999—),女,河北廊坊人,硕士研究生,CCF会员,主要研究方向:数据分析。
  • 基金资助:
    国家自然科学基金资助项目(U20A20179)

Federated learning algorithm for personalization and fairness

Hongyang ZHANG1,2,3, Shufen ZHANG1,2,4(), Zheng GU1,2,3   

  1. 1.College of Sciences,North China University of Science and Technology,Tangshan Hebei 063210,China
    2.Hebei Provincial Key Laboratory of Data Science and Application (North China University of Science and Technology),Tangshan Hebei 063210,China
    3.Tangshan Key Laboratory of Data Science (North China University of Science and Technology),Tangshan Hebei 063210,China
    4.Tangshan Key Laboratory of Big Data Security and Intelligent Computing (North China University of Science and Technology),Tangshan Hebei 063210,China
  • Received:2024-07-05 Revised:2024-09-25 Accepted:2024-09-29 Online:2025-07-10 Published:2025-07-10
  • Contact: Shufen ZHANG
  • About author:ZHANG Hongyang, born in 1999, M. S. candidate. His research interests include data security, privacy protection.
    ZHANG Shufen, born in 1972, M. S., professor. Her research interests include cloud computing, data security, privacy protection.
    GU Zheng, born in 1999, M. S. candidate. Her research interests include data analysis.
  • Supported by:
    National Natural Science Foundation of China(U20A20179)

摘要:

作为一种分布式优化范式,联邦学习(FL)允许大量资源有限的客户端节点在不共享数据时协同训练模型。然而,传统联邦学习算法,如FedAvg,通常未充分考虑公平性的问题。在实际场景中,数据分布通常具备高度异构性,常规的聚合操作可能会使模型对某些客户端产生偏见,导致全局模型在客户端本地的性能分布出现巨大差异。针对这一问题,提出一种面向个性化与公平性的联邦学习FedPF(Federated learning for Personalization and Fairness)算法。FedPF旨在有效减少联邦学习中低效的聚合行为,并通过寻找全局模型与本地模型的相关性,在客户端之间分配个性化模型,从而在保证全局模型性能的同时,使客户端本地性能分布更均衡。将FedPF在Synthetic、MNIST以及CIFAR10数据集上进行实验和性能分析,并与FedProx、q-FedAvg和FedAvg这3种联邦学习算法进行对比。实验结果表明,FedPF在有效性和公平性上均得到了有效提升。

关键词: 联邦学习, 公平, 个性化, 异构数据, 客户端选择

Abstract:

As a distributed optimization paradigm, Federated Learning (FL) enables a large number of resource-constrained client nodes to train models collaboratively without sharing the data. However, traditional federated learning algorithms, such as fedAvg, often fail to address fairness issues adequately. In practical scenarios, data distributions are highly heterogeneous typically, and conventional aggregation operations may introduce model biases towards certain clients, resulting in significant local performance disparities across clients of the global model. To tackle this challenge, a federated learning algorithm for personalization and fairness named FedPF (Federated learning for Personalization and Fairness) was proposed. FedPF aims to reduce inefficient aggregation behaviors in federated learning effectively, and distribute personalized models among clients by exploring the correlations between the global model and local models, thereby ensuring a balanced performance distribution among clients while maintaining performance of the global model. FedPF was evaluated and analyzed on Synthetic, MNIST, and CIFAR10 datasets, and was compared with three federated learning algorithms: FedProx, q-FedAvg, and FedAvg. Experimental results demonstrate that FedPF achieves notable improvements in both effectiveness and fairness.

Key words: federated learning, fairness, personalization, heterogeneous data, client selection

中图分类号: