About this document

FedWSQ: Optimizing Federated Learning by momoland1069 is a document available to read on EtoBox.

The document presents FedWSQ, a novel federated learning framework that combines weight standardization (WS) and distribution-aware non-uniform quantization (DANUQ) to address challenges like data heterogeneity and communication constraints. FedWSQ enhances model robustness and communication efficiency, achieving superior performance across various federated learning settings, even under extreme data heterogeneity and low-bit communication conditions. Extensive experiments demonstrate that FedWSQ consistent

Author

momoland1069

Language

EN