Subgraph Generation for Generalizing on Out-of-Distribution Links

2025-07-15

AI Summary

Graph neural networks (GNNs) suffer from poor generalization in out-of-distribution (OOD) link prediction, while graph generation models (GGMs) remain underexploited for this task. Method: We propose FLEX—a novel framework that introduces GGMs to OOD link prediction for the first time. It performs structure-conditioned subgraph generation to enable expert-free data augmentation; designs an adversarial co-training mechanism between a graph autoencoder and a GNN to enforce structural consistency across distributions; and jointly optimizes generative and discriminative modules to enhance structural distributional robustness. Contribution/Results: Evaluated on multiple synthetic and real-world OOD benchmarks, FLEX consistently outperforms state-of-the-art methods, demonstrating both the effectiveness and generalizability of generative structural augmentation for OOD link prediction.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Natural Language Processing: Generation

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems

Abstract

Graphs Neural Networks (GNNs) demonstrate high-performance on the link prediction (LP) task. However, these models often rely on all dataset samples being drawn from the same distribution. In addition, graph generative models (GGMs) show a pronounced ability to generate novel output graphs. Despite this, GGM applications remain largely limited to domain-specific tasks. To bridge this gap, we propose FLEX as a GGM framework which leverages two mechanism: (1) structurally-conditioned graph generation, and (2) adversarial co-training between an auto-encoder and GNN. As such, FLEX ensures structural-alignment between sample distributions to enhance link-prediction performance in out-of-distribution (OOD) scenarios. Notably, FLEX does not require expert knowledge to function in different OOD scenarios. Numerous experiments are conducted in synthetic and real-world OOD settings to demonstrate FLEX's performance-enhancing ability, with further analysis for understanding the effects of graph data augmentation on link structures. The source code is available here: https://github.com/revolins/FlexOOD.

Problem

• Improving link prediction in out-of-distribution graphs
• Bridging graph generative models and domain-specific limitations
• Enhancing structural alignment without expert OOD knowledge

Innovation

• Structurally-conditioned graph generation mechanism
• Adversarial co-training of auto-encoder and GNN
• Expert-free OOD adaptation for link prediction