SAOT: Self-Supervised Continual Graph Learning with Structure-Aware Optimal Transport

2026-06-30

AI Summary

This work addresses the issue of isolated node optimization disrupting global relational structures in self-supervised continual graph learning. To this end, it proposes the first structure-aware optimal transport framework that enables continual learning over graph-structured tasks without labeled data. The method explicitly models and preserves global node correspondences across sequential tasks through optimal transport, while integrating a knowledge distillation mechanism to facilitate effective knowledge transfer. Extensive experiments on four benchmark datasets demonstrate that the proposed approach significantly outperforms existing methods, achieving absolute gains of up to 5% in average accuracy on CoraFull-CL and over 15% on Products-CL.

Abstract

Self-supervised Continual Graph Learning (CGL) aims to successively learn from a graph sequence with different tasks without label supervision - a paradigm that has attracted widespread attention. Most existing self-supervised CGL methods rely on instance-level consistency objectives that enforce stability of individual node (or node-pair) embeddings. Due to optimizing nodes in isolation, these methods fail to maintain global relational structure, causing inter-node correspondences to progressively distort under continual learning. To this end, we propose a novel Structure-Aware Optimal Transport (SAOT) framework that explicitly captures and preserves relational structure within graph representations across sequential tasks. Specifically, SAOT leverages optimal transport theory to capture global inter-node correspondences, thereby facilitating and enhancing graph representation learning. Simultaneously, SAOT incorporates a cross-task knowledge distillation mechanism to preserve the previous structural knowledge. Extensive experiments on four CGL benchmark datasets demonstrate that SAOT outperforms existing self-supervised baselines. In particular, SAOT achieves significant performance gains, improving average accuracy by up to 5% on CoraFull-CL and over 15% on Products-CL compared with state-of-the-art methods in the Class-IL setting.

Problem

• Continual Graph Learning
• Self-Supervised Learning
• Relational Structure Preservation
• Optimal Transport
• Graph Representation Learning

Innovation

• Self-Supervised Continual Graph Learning
• Optimal Transport
• Structure Preservation
• Knowledge Distillation
• Graph Representation Learning