This is a page that curates AI-related papers published worldwide. All content here is summarized using Google Gemini and operated on a non-profit basis. Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.
Zihan Tan, Suyuan Huang, Guancheng Wan, Wenke Huang, He Li, Mang Ye
Outline
In this paper, we present Federated Graph Learning (FGL), which combines the privacy-preserving capabilities of Federated Learning (FL) with the powerful graph modeling capabilities of Graph Neural Networks (GNNs). Existing studies only address subgraph FL from a structural perspective, overlooking the propagation of graph signals in spatial and spectral domains. From a spatial perspective, subgraph FL causes edge disconnection between clients, which leads to label signal disruption and class knowledge degradation of global GNNs. From a spectral perspective, spectral heterogeneity causes signal frequency mismatch between subgraphs, which causes local GNNs to overfit to local signal propagation schemes. As a result, spectral client drift occurs, which deteriorates global generalization. To address these issues, we propose a global knowledge repository that mitigates label signal disruption and a frequency alignment that addresses spectral client drift. We construct the S2FGL framework by combining spatial and spectral strategies. We demonstrate the superiority of S2FGL through extensive experiments on multiple datasets. The code can be found at https://github.com/Wonder7racer/S2FGL.git .
