Graph Self-Supervised Learning with Learnable Structural and Positional Encodings
This work addresses limitations in graph representation learning for tasks like graph classification, advancing GSSL's ability to distinguish graphs with similar local structures but different global topologies, though it appears incremental as it builds on existing GNN and self-supervised learning methods.
The paper tackles the problem of Graph Self-Supervised Learning (GSSL) struggling to capture complex structural properties by introducing GenHopNet, a GNN framework with a k-hop message-passing scheme and a structural- and positional-aware GSSL framework, which consistently outperforms existing approaches in graph classification experiments.
Traditional Graph Self-Supervised Learning (GSSL) struggles to capture complex structural properties well. This limitation stems from two main factors: (1) the inadequacy of conventional Graph Neural Networks (GNNs) in representing sophisticated topological features, and (2) the focus of self-supervised learning solely on final graph representations. To address these issues, we introduce \emph{GenHopNet}, a GNN framework that integrates a $k$-hop message-passing scheme, enhancing its ability to capture local structural information without explicit substructure extraction. We theoretically demonstrate that \emph{GenHopNet} surpasses the expressiveness of the classical Weisfeiler-Lehman (WL) test for graph isomorphism. Furthermore, we propose a structural- and positional-aware GSSL framework that incorporates topological information throughout the learning process. This approach enables the learning of representations that are both sensitive to graph topology and invariant to specific structural and feature augmentations. Comprehensive experiments on graph classification datasets, including those designed to test structural sensitivity, show that our method consistently outperforms the existing approaches and maintains computational efficiency. Our work significantly advances GSSL's capability in distinguishing graphs with similar local structures but different global topologies.