Taming Cross-Domain Representation Variance in Federated Prototype Learning with Heterogeneous Data Domains

arXiv:2403.09048v1 Announce Type: new
Abstract: Federated learning (FL) allows collaborative machine learning training without sharing private data. While most FL methods assume identical data domains across clients, real-world scenarios often involve heterogeneous data domains. Federated Prototype Learning (FedPL) addresses this issue, using mean feature vectors as prototypes to enhance model generalization. However, existing FedPL methods create the same number of prototypes for each client, leading to cross-domain performance gaps and disparities for clients with varied data distributions. To mitigate cross-domain …

abstract arxiv collaborative cs.cv cs.lg data domain domains feature federated learning issue machine machine learning mean private data representation training type variance vectors world