Back to Author Console Empowering GNNs for Domain Adaptation via Denoising Target Graph

2025-12-05

AI Summary

In graph domain adaptation, structural domain shift degrades the generalization of Graph Neural Networks (GNNs) on target-graph node classification. To address this, we propose GraphDeT—a framework that jointly leverages source/target graph structures and source labels via end-to-end training, with edge denoising as an auxiliary self-supervised task. This design mitigates negative transfer induced by cross-domain structural inconsistency and theoretically yields a tighter graph generalization bound. Extensive experiments on real-world transfer scenarios—including temporal evolution and geographical regions—demonstrate that GraphDeT significantly outperforms state-of-the-art methods, confirming its effectiveness and robustness. Our key contribution lies in the first formulation of edge denoising as a structure-aware self-supervised signal, coupled with its synergistic optimization alongside supervised learning.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community
• Computer Vision: Adversarial Attacks & Robustness

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies
• Search and Retrieval-Augmented AI: Vertical and domain-specific search

Abstract

We explore the node classification task in the context of graph domain adaptation, which uses both source and target graph structures along with source labels to enhance the generalization capabilities of Graph Neural Networks (GNNs) on target graphs. Structure domain shifts frequently occur, especially when graph data are collected at different times or from varying areas, resulting in poor performance of GNNs on target graphs. Surprisingly, we find that simply incorporating an auxiliary loss function for denoising graph edges on target graphs can be extremely effective in enhancing GNN performance on target graphs. Based on this insight, we propose our framework, GraphDeT, a framework that integrates this auxiliary edge task into GNN training for node classification under domain adaptation. Our theoretical analysis connects this auxiliary edge task to the graph generalization bound with -distance, demonstrating such auxiliary task can imposes a constraint which tightens the bound and thereby improves generalization. The experimental results demonstrate superior performance compared to the existing baselines in handling both time and regional domain graph shifts.

Problem

• Enhancing GNN generalization for node classification across domain shifts
• Addressing performance drop due to structural differences in source and target graphs
• Proposing a denoising edge auxiliary task to tighten graph generalization bounds

Innovation

• Denoising target graph edges via auxiliary loss
• Integrating edge task into GNN training for adaptation
• Theoretical link to graph generalization bound improvement