This repository contains the code for our paper CosPer: Fair and robust personalized federated learning with heterogeneous data adaptation
Abstract: This paper propose an innovative personalized federated learning (PFL) framework named CosPer to address these competing constraints. The framework employs an adaptive local aggregation mechanism, where the aggregation weight is contingent upon the cosine similarity between the global gradient and each local gradient. This mechanism can construct a personalized model for each client, suited to different degrees of data heterogeneity.
- Code base: PFL-Non-IID (PyTorch Version)