Graph Augmentation for Cross Graph Domain Generalization

2025-02-25

AI Summary

This paper addresses domain generalization in cross-graph node classification—leveraging labeled nodes from source graphs to classify unlabeled nodes in out-of-distribution (OOD) target graphs. We propose a novel graph-structured data augmentation method that departs from conventional model-centric optimization paradigms. Our approach employs a joint augmentation strategy: first pruning low-weight edges to reinforce a robust topological skeleton, then generating cross-graph invariant semantic edges guided by node feature clustering. By explicitly modeling structural invariance, the method significantly enhances the robustness of Graph Neural Networks (GNNs) to distributional shifts. Evaluated on multiple OOD citation network benchmarks, our method consistently outperforms existing graph augmentation baselines and achieves state-of-the-art performance.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community
• Search and Optimization: Distributed Search

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

Cross-graph node classification, utilizing the abundant labeled nodes from one graph to help classify unlabeled nodes in another graph, can be viewed as a domain generalization problem of graph neural networks (GNNs) due to the structure shift commonly appearing among various graphs. Nevertheless, current endeavors for cross-graph node classification mainly focus on model training. Data augmentation approaches, a simple and easy-to-implement domain generalization technique, remain under-explored. In this paper, we develop a new graph structure augmentation for the crossgraph domain generalization problem. Specifically, low-weight edgedropping is applied to remove potential noise edges that may hinder the generalization ability of GNNs, stimulating the GNNs to capture the essential invariant information underlying different structures. Meanwhile, clustering-based edge-adding is proposed to generate invariant structures based on the node features from the same distribution. Consequently, with these augmentation techniques, the GNNs can maintain the domain invariant structure information that can improve the generalization ability. The experiments on out-ofdistribution citation network datasets verify our method achieves state-of-the-art performance among conventional augmentations.

Problem

• Enhances cross-graph node classification
• Proposes graph structure augmentation techniques
• Improves generalization in graph neural networks

Innovation

• Low-weight edge-dropping technique
• Clustering-based edge-adding method
• Enhances GNN generalization ability