
FedCRL leverages contrastive representation learning on shared representations between clients and the server to mitigate the challenges of label distribution skew and data scarcity in federated learning scenarios, thereby enhancing personalized model performance across heterogeneous clients.


coremsg

federated-contrastive-representation-learning-enhancing-personalized-federated-learning-for-label-heterogeneity-in-non-iid-data


Federated Contrastive Representation Learning: Enhancing Personalized Federated Learning for Label Heterogeneity in Non-IID Data
