Abstract:Graph contrastive learning (GCL) is a promising direction toward alleviating the label dependence, poor generalization and weak robustness of graph neural networks, learning representations with invariance, and discriminability by solving pretasks. The pretasks are mainly built on mutual information estimation, which requires data augmentation to construct positive samples with similar semantics to learn invariant signals and negative samples with dissimilar semantics to empower representation discriminability. However, an appropriate data augmentation configuration depends heavily on lots of empirical trials such as choosing the compositions of data augmentation techniques and the corresponding hyperparameter settings. We propose an augmentation-free GCL method, invariant-discriminative GCL (iGCL), that does not intrinsically require negative samples. iGCL designs the invariant-discriminative loss (ID loss) to learn invariant and discriminative representations. On the one hand, ID loss learns invariant signals by directly minimizing the mean square error (MSE) between the target samples and positive samples in the representation space. On the other hand, ID loss ensures that the representations are discriminative by an orthonormal constraint forcing the different dimensions of representations to be independent of each other. This prevents representations from collapsing to a point or subspace. Our theoretical analysis explains the effectiveness of ID loss from the perspectives of the redundancy reduction criterion, canonical correlation analysis (CCA), and information bottleneck (IB) principle. The experimental results demonstrate that iGCL outperforms all baselines on five node classification benchmark datasets. iGCL also shows superior performance for different label ratios and is capable of resisting graph attacks, which indicates that iGCL has excellent generalization and robustness. The source code is available at https://github.com/lehaifeng/ T-GCN/tree/master/iGCL.

Spectral-Aware Augmentation for Enhanced Graph Representation Learning

Spectral Augmentation for Self-Supervised Learning on Graphs

Spectral Augmentations for Graph Contrastive Learning

SimGRACE: A Simple Framework for Graph Contrastive Learning Without Data Augmentation

Heterogeneous Graph Contrastive Learning with Spectral Augmentation

Revisiting Graph Contrastive Learning from the Perspective of Graph Spectrum

Augmentation-Free Graph Contrastive Learning with Performance Guarantee

Graph Contrastive Learning with Personalized Augmentation

Rethinking Spectral Augmentation for Contrast-based Graph Self-Supervised Learning

LSPAN: Spectrally Localized Augmentation for Graph Consistency Learning

Graph Contrastive Learning with Adaptive Augmentation

Boosting Graph Contrastive Learning via Graph Contrastive Saliency

Graph Augmentation for Recommendation

Community-Invariant Graph Contrastive Learning

Augmentations in Graph Contrastive Learning: Current Methodological Flaws & Towards Better Practices

Augmentation-Free Graph Contrastive Learning of Invariant-Discriminative Representations

Graph Contrastive Learning with Reinforcement Augmentation

MA-GCL: Model Augmentation Tricks for Graph Contrastive Learning

Graph Contrastive Learning with Reinforced Augmentation

Graph Contrastive Learning with Cohesive Subgraph Awareness

SpeGCL: Self-supervised Graph Spectrum Contrastive Learning without Positive Samples