Article Details

Abstract

This paper aims to systematically evaluate the performance of representative personalized federated learning algorithms under data heterogeneity. Based on the lightweight federated learning platform LightFL, ten representative personalized algorithms are implemented and tested on the MNIST and CIFAR-10 datasets under both IID and Dirichlet Non-IID data partitioning scenarios. The generalization performance, personalization capability, and relative performance improvement of each algorithm are experimentally compared. The results show that in simple tasks, most algorithms remain stable under Non-IID settings, while in complex tasks, FedRep, FedAMP, and FedALA exhibit stronger personalization performance. Therefore, lightweight algorithms such as FedRep and FedBN are recommended for simple tasks; FedRep, FedAMP, and FedALA are preferred for complex tasks; and FedRep and FedALA are robust choices when task complexity is unknown. This study provides practical guidance for selecting personalized federated learning algorithms in real-world applications.

Keywords

personalized federated learning; algorithm evaluation; independent and identically distributed; non-independent and identically distributed

References

1. [1] Mcmahan B，Moore E，Ramage D，et al. Communication-Efficient Learning of Deep Networks from Decentralized Data [C]//Proceedings of the 20th International Conference on Artificial Intelligence and Statistics，2017：1273-1282.
2. [2] Liu Ruoxuan, Gao Linghang, Xie Guoyun, etc Design and Implementation of Lightweight Federated Learning Platform LightFL [J]. Modern Information Technology, 2025, 9 (6): 71-74
3. [3] Arivazhagan M G，Aggarwal V，Singh A K，et al. Federated Learning with Personalization Layers [PP/OL].arXiv:1912.00818 [cs.LG].https://doi.org/10.48550/arXiv.1912.00818.
4. [4] Collins L，Hassani H，Mokhtari A，et al. Exploiting Shared Representations for Personalized Federated Learning [C]//Proceedings of the 38th International Conference on Machine Learning，2021：2089-2099.
5. [5] Li X X，Jiang M R，Zhang X F，et al. FedBN: Federated Learning on Non-IID Features via Local Batch Normalization [C]//International Conference on Learning Representations，2021：1-27.
6. [6] Dinh C T，Tran N H，Nguyen T D. Personalized Federated Learning with Moreau Envelopes [C]//Proceedings of the 34th International Conference on Neural Information Processing Systems，2020：21394-21405.
7. [7] Diao E，Ding J，Tarokh V. HeteroFL: Computation and Communication Efficient Federated Learning for Heterogeneous Clients [PP/OL].arXiv:2010.01264 [cs.LG].https://doi.org/10.48550/arXiv.2010.01264.
8. [8] Zhu Z D，Hong J Y，Zhou J Y. Data-Free Knowledge Distillation for Heterogeneous Federated Learning [C]//Proceedings of machine learning research，2021：12878-12889.
9. [9] Zhang M，Sapra K，Fidler S，et al. Personalized Federated Learning with First Order Model Optimization [PP/OL].arXiv:2012.08565 [cs.LG].https://doi.org/10.48550/arXiv.2012.08565.
10. [10] Huang Y T，Chu L Y，Zhou Z R，et al. Personalized Cross-Silo Federated Learning on Non-Iid Data [C]//Proceedings of the AAAI Conference on Artificial Intelligence，2021，35（9）：7865-7873.
11. [11] Zhang J，Hua Y，Wang H，et al. FedALA: Adaptive Local Aggregation for Personalized Federated Learning [J].Proceedings of the AAAI conference on artificial intelligence，2023，37（9）：11237-11244.
12. [12] Xu J，Tong X Y，Huang S L. Personalized Federated Learning with Feature Alignment and Classifier Collaboration [PP/OL].arXiv.https://doi.org/10.48550/arXiv.2306.11867.

Funding

2025 Beijing University of Information Science and Technology Innovation and Entrepreneurship Training Program for College Students (X202511232476); Support Program for Young Backbone Teachers of Beijing University of Information Science and Technology in 2024 (YBT202450); Development of an AI based weather forecasting model for the Qinghai Tibet Plateau region (S2426030); 2026 Beijing University of Information Science and Technology Undergraduate Teaching Internal Project "Promoting the Classification Development of Universities - High Quality Curriculum Construction and Reform Project of the School of Science" (5112623903)

About the authors

Chen Jianchu (2005-), male, Han ethnicity, from Anshun, Guizhou, undergraduate student, research interests include federated learning; Chen Youxi (2005-), female, Han ethnicity, from Yongchuan, Chongqing, undergraduate student, research interests include federated learning; Ma Wenxuan (2005-), male, Hui ethnicity, from Wuzhong, Ningxia, undergraduate student, research interests include federated learning; Kong Ziyu (2005-), male, Han ethnicity, from Mianyang, Sichuan, undergraduate student, research interests include federated learning; Li Shangqing (2005-), male, Han ethnicity, from Xinxiang, Henan, undergraduate student, research interests include federated learning; Corresponding author: Wang Can (1990-), male, Han ethnicity, from Jining, Shandong, associate professor, doctoral degree, research interests include federated learning.