SpecGradFilter: A Spectral Gradient Filtering Framework for Taming Federated Heterogeneity
This work revisits client drift from a novel frequency-domain perspective and uncovers a critical Spectral Bias of Drift: inter-client gradient divergence is predominantly concentrated in low-frequency components which encode client-specific distributional shifts, while high-frequency components representing fine-grained features remain relatively consistent.
Liyang Yuan, Yibo Yang, Dandan Guo et al.
· 0 citations