Mitigating Heterogeneity-Induced Drift in Hierarchical Sign-Based Federated Learning
Researchers have developed a new framework for hierarchical federated learning that addresses the issue of data heterogeneity across different clusters. The proposed DC-HierSignSGD algorithm uses binary sign-based stochastic gradient descent with a cloud-assisted correction mechanism to mitigate bias and improve model stability and accuracy. This approach aims to achieve performance comparable to full-precision methods while significantly reducing communication overhead, particularly in large-scale Internet of Things systems. AI