Communication-efficient Federated Graph Classification via Generative Diffusion Modeling
Xiuling Wang, Xin Huang, Haibo Hu, Jianliang Xu
TL;DR
CeFGC tackles the dual challenges of communication overhead and non-IID data in federated graph learning by introducing a three-round FGNN framework that uses graph generative diffusion models (GGDMs) to share synthetic data rather than frequent parameter updates. It offers two GGDM training variants (basic per-class models and an advanced class-label channel) and a four-phase workflow that includes encryption-based privacy safeguards and a single FedAvg-style global aggregation. The approach achieves substantial reductions in communication rounds and data transfer while maintaining or improving graph classification performance across diverse non-IID settings, with strong empirical results on five real datasets. The work also demonstrates extensions to other graph learning tasks and discusses privacy enhancements, offering a practical pathway for scalable, privacy-aware federated graph learning in non-IID environments.
Abstract
Graph Neural Networks (GNNs) unlock new ways of learning from graph-structured data, proving highly effective in capturing complex relationships and patterns. Federated GNNs (FGNNs) have emerged as a prominent distributed learning paradigm for training GNNs over decentralized data. However, FGNNs face two significant challenges: high communication overhead from multiple rounds of parameter exchanges and non-IID data characteristics across clients. To address these issues, we introduce CeFGC, a novel FGNN paradigm that facilitates efficient GNN training over non-IID data by limiting communication between the server and clients to three rounds only. The core idea of CeFGC is to leverage generative diffusion models to minimize direct client-server communication. Each client trains a generative diffusion model that captures its local graph distribution and shares this model with the server, which then redistributes it back to all clients. Using these generative models, clients generate synthetic graphs combined with their local graphs to train local GNN models. Finally, clients upload their model weights to the server for aggregation into a global GNN model. We theoretically analyze the I/O complexity of communication volume to show that CeFGC reduces to a constant of three communication rounds only. Extensive experiments on several real graph datasets demonstrate the effectiveness and efficiency of CeFGC against state-of-the-art competitors, reflecting our superior performance on non-IID graphs by aligning local and global model objectives and enriching the training set with diverse graphs.