Robust method helps multiple devices learn safely even with bad data
Robust Decentralized Personalized Federated Learning via Prediction-Constrained Neighborhood Collaboration
Machine LearningCryptography and SecurityDistributed, Parallel, and Cluster Computing
Summary
When many devices work together to learn from data, some may send wrong or harmful information that can confuse the learning process. The authors propose a method called R-DPFL that helps each device check its neighbors' updates carefully by predicting what good updates should look like based on past behavior. This helps them ignore or reduce the impact of bad or misleading updates without needing everyone to fully agree. They tested their method on a common image dataset and found it works better than existing approaches in tricky situations where some devices act badly or data is very different across devices.
federated learningdecentralized learningByzantine attackspersonalized modelsneighborhood collaborationupdate aggregationrobustnessmodel convergenceCIFAR-10adversarial settings
Authors
Xiao Ma, Hong Shen, Hui Tian, Wenqi Lyu, Wei Ke
Abstract
This paper proposes a robust decentralized personalized federated learning method R-DPFL, that enables clients to reduce the impact of Byzantine attacks via robust neighborhood direction estimation and history-based update trend prediction, rather than purely aggregating client models as in the existing work. In R-DPFL, each client first computes the current-round model update by aggregating the received neighborhood update vectors. It then predicts what this update should be based on its historical values and local model changes. Finally, R-DPFL computes the difference between these two quantities, adaptively clips this difference, and adds it to the local update. We prove convergence of the learning process through rigorous analysis and show that honest clients maintain stable personalized descent dynamics under Byzantine neighbor perturbations without requiring consensus among neighboring models. Extensive experiments on CIFAR-10 demonstrate that RDPFL consistently outperforms state-of-the-art decentralized and personalized federated learning baselines under heterogeneous and adversarial settings.