AS-GCL: Asymmetric Spectral Augmentation on Graph Contrastive Learning

TL;DR

This work proposes a novel paradigm called AS-GCL that incorporates asymmetric spectral augmentation for graph contrastive learning and is the first to encode augmentation views of the spectral domain using asymmetric encoders.

Abstract

Graph Contrastive Learning (GCL) has emerged as the foremost approach for self-supervised learning on graph-structured data. GCL reduces reliance on labeled data by learning robust representations from various augmented views. However, existing GCL methods typically depend on consistent stochastic augmentations, which overlook their impact on the intrinsic structure of the spectral domain, thereby limiting the model's ability to generalize effectively. To address these limitations, we propose a novel paradigm called AS-GCL that incorporates asymmetric spectral augmentation for graph contrastive learning. A typical GCL framework consists of three key components: graph data augmentation, view encoding, and contrastive loss. Our method introduces significant enhancements to each of these components. Specifically, for data augmentation, we apply spectral-based augmentation to minimize spectral variations, strengthen structural invariance, and reduce noise. With respect to encoding, we employ parameter-sharing encoders with distinct diffusion operators to generate diverse, noise-resistant graph views. For contrastive loss, we introduce an upper-bound loss function that promotes generalization by maintaining a balanced distribution of intra- and inter-class distance. To our knowledge, we are the first to encode augmentation views of the spectral domain using asymmetric encoders. Extensive experiments on eight benchmark datasets across various node-level tasks demonstrate the advantages of the proposed method.

Figures (7)

• Figure 1: The general framework of the proposed AS-GCL method. First, topology augmentation is optimized by generating paired augmented views $\mathcal{G}_1$ and $\mathcal{G}_2$ through spectral variation minimization. Then, encoders with shared parameters but different diffusion operators are used to generate representations for different views ($\boldsymbol{H}^1$, and $\boldsymbol{H}^2$). Finally, a contrastive loss function is introduced to reduce intraclass variation and increase interclass contrast.
• Figure 2: Training loss for different numbers of rounds.
• Figure 3: Frobenius distance between the symmetric normalized Laplace matrices of the decomposition of the original and augmented graphs.
• Figure 4: Node classification accuracy with different edge perturbation ratios $\epsilon$.
• Figure 5: Node classification accuracy with different diffusion layers $k$ in asymmetric encoders.

Theorems & Definitions (3)

• Remark 1
• Remark 2
• Remark 3