Towards Federated Long-Tailed Graph Learning: An Energy-Guided Dual Decoupling Approach

2026-06-23

AI Summary

This work addresses the challenge in federated graph learning where long-tailed class distributions cause the global model to favor majority classes, leading minority-class nodes to be overwhelmed within heterophilous neighborhoods. To mitigate this, the authors propose FedEPD, a novel framework that introduces, for the first time, an energy-guided dual disentanglement mechanism. Specifically, it employs distribution-aware Dirichlet energy pruning to remove heterophilous edges, thereby separating structural noise from semantic information. Additionally, global prototypes are extracted from topologically central nodes and injected into local representations via spatial low-pass filtering. A two-stage alternating optimization strategy is adopted to simultaneously preserve majority-class decision boundaries and enhance minority-class performance. Evaluated on multiple long-tailed graph benchmarks, FedEPD achieves state-of-the-art results, with accuracy improvements up to 4.97% and Macro-F1 gains as high as 5.48%.

Abstract

Federated Graph Learning facilitates collaborative graph modeling across distributed clients while preserving data privacy. However, real-world data categories frequently exhibit long-tailed distributions. Such statistical scarcity severely degrades performance in two ways: it biases the global model toward majority classes, and it structurally isolates minority nodes by submerging them in heterophilic, head-dominated neighborhoods. While existing methods attempt topology-agnostic statistical compensations, they often fail under data scarcity. Instead of recovering tail nodes, they overfit the structural noise from adjacent dominant classes, leading to representation degradation. To address these limitations, we propose FedEPD, a framework built on a dual decoupling paradigm that separates topological purification from semantic recalibration. Specifically, FedEPD utilizes distribution-aware Dirichlet energy pruning to filter spatial heterophilic edges. It then overcomes Non-IID distribution shifts by extracting robust global prototypes from topologically central nodes, which are incorporated into local representations via a spatial low-pass prototype injection. Furthermore, a two stage alternating optimization strategy strictly protects majority decision boundaries while improving minority accuracy. Extensive experiments demonstrate that FedEPD achieves state-of-the-art performance across diverse long-tailed benchmarks, yielding absolute improvements of up to 4.97% in Accuracy and 5.48% in Macro-F1.

Problem

• Federated Graph Learning
• Long-Tailed Distribution
• Data Scarcity
• Structural Heterophily
• Minority Nodes

Innovation

• Federated Graph Learning
• Long-Tailed Distribution
• Dual Decoupling
• Dirichlet Energy Pruning
• Prototype Injection