Variational Perturbation Personalized Federated Learning via Prior-Posterior Distance

Hefeng Zhou; Yuanbin Wang; Jun Wang; Jiong Lou; Wugedele Bao; Chentao Wu; Jie Li

ICASSP 2025accepted0 citations

Variational Perturbation Personalized Federated Learning via Prior-Posterior Distance

Hefeng Zhou, Yuanbin Wang, Jun Wang, Jiong Lou, Wugedele Bao, Chentao Wu, Jie Li

Abstract

Personalized Federated Learning (pFL) mitigates the impact of statistical heterogeneity on FL architecture to some extent by allowing participants to use personalized models based on local data distributions. The existing pFL methods optimize from the perspective of model structure, attempting to adopt strategies that maintain model processing or quickly adapt to local data distribution capabilities. Our proposed method draws inspiration from the concept of variational inference, guiding model updates by comparing prior and posterior data distributions, and innovatively applying model variational perturbations to improve robustness. Finally, we conducted multidimensional experiments and the results show that our method outperforms the current baseline. Code: https://github.com/RezinChow/VPFL.

BibTeX

@inproceedings{icassp2025_variationalpertu,
  title = {Variational Perturbation Personalized Federated Learning via Prior-Posterior Distance},
  author = {Hefeng Zhou and Yuanbin Wang and Jun Wang and Jiong Lou and Wugedele Bao and Chentao Wu and Jie Li},
  booktitle = {ICASSP 2025},
  year = {2025}
}