FedWeight: mitigating covariate shift of federated learning on electronic health records data through patients re-weighting
Federated learning (FL) enables collaborative analysis of decentralized medical data while preserving patient privacy. However, the covariate shift from demographic and clinical differences can reduce model generalizability. We propose FedWeight, a novel FL framework…
Continue Reading