FEMNIST

Federated Learning uses spectral entropy for data-free client contribution estimation

Researchers have developed a novel method for estimating client contributions in Federated Learning without requiring access to client data. This approach utilizes the spectral entropy of final-layer updates to measure …