Graph Contrastive Learning via Spectral Graph Alignment
This addresses a specific bottleneck in graph contrastive learning for researchers in graph representation learning, though it appears incremental as it builds on existing contrastive frameworks.
The paper tackles the problem of controlling global structure in graph contrastive learning by introducing SpecMatch-CL, a loss function that aligns view-specific graph-of-graphs via normalized Laplacian differences. The method achieves new state-of-the-art results on eight TU benchmarks and shows consistent gains in transfer learning on PPI-306K and ZINC 2M datasets.
Given augmented views of each input graph, contrastive learning methods (e.g., InfoNCE) optimize pairwise alignment of graph embeddings across views while providing no mechanism to control the global structure of the view specific graph-of-graphs built from these embeddings. We introduce SpecMatch-CL, a novel loss function that aligns the view specific graph-of-graphs by minimizing the difference between their normalized Laplacians. Theoretically, we show that under certain assumptions, the difference between normalized Laplacians provides an upper bound not only for the difference between the ideal Perfect Alignment contrastive loss and the current loss, but also for the Uniformly loss. Empirically, SpecMatch-CL establishes new state of the art on eight TU benchmarks under unsupervised learning and semi-supervised learning at low label rates, and yields consistent gains in transfer learning on PPI-306K and ZINC 2M datasets.