Abstract:Graph Contrastive Learning (GCL) has emerged as a promising approach in the realm of graph self-supervised learning. Prevailing GCL methods mainly derive from the principles of contrastive learning in the field of computer vision: modeling invariance by specifying absolutely similar pairs. However, when applied to graph data, this paradigm encounters two significant limitations: (1) the validity of the generated views cannot be guaranteed: graph perturbation may produce invalid views against semantics and intrinsic topology of graph data; (2) specifying absolutely similar pairs in the graph views is unreliable: for abstract and non-Euclidean graph data, it is difficult for humans to decide the absolute similarity and dissimilarity intuitively. Despite the notable performance of current GCL methods, these challenges necessitate a reevaluation: Could GCL be more effectively tailored to the intrinsic properties of graphs, rather than merely adopting principles from computer vision? In response to this query, we propose a novel paradigm, Graph Soft-Contrastive Learning (GSCL). This approach facilitates GCL via neighborhood ranking, avoiding the need to specify absolutely similar pairs. GSCL leverages the underlying graph characteristic of diminishing label consistency, asserting that nodes that are closer in the graph are overall more similar than far-distant nodes. Within the GSCL framework, we introduce pairwise and listwise gated ranking InfoNCE loss functions to effectively preserve the relative similarity ranking within neighborhoods. Moreover, as the neighborhood size exponentially expands with more hops considered, we propose neighborhood sampling strategies to improve learning efficiency. Our extensive empirical results across 11 commonly used graph datasets-including 8 homophily graphs and 3 heterophily graphs-demonstrate GSCL's superior performance compared to 20 SOTA GCL methods.

NNC-GCN: Neighbours-to-Neighbours Contrastive Graph Convolutional Network for Semi-Supervised Classification

G-Censor: Graph Contrastive Learning with Task-Oriented Counterfactual Views

Contrastive Multi-graph Learning with Neighbor Hierarchical Sifting for Semi-supervised Text Classification

CC-GNN: A Clustering Contrastive Learning Network for Graph Semi-Supervised Learning

Neighbor Contrastive Learning on Learnable Graph Augmentation

Contrastive Graph Representation Learning with Adversarial Cross-view Reconstruction and Information Bottleneck

Local Structure-aware Graph Contrastive Representation Learning

Supervised contrastive learning for graph representation enhancement

Single-View Graph Contrastive Learning with Soft Neighborhood Awareness

Graph Soft-Contrastive Learning via Neighborhood Ranking

Contrastive learning of graphs under label noise

Perturbation-augmented Graph Convolutional Networks: A Graph Contrastive Learning architecture for effective node classification tasks

Localized Contrastive Learning on Graphs

Enhancing Graph Contrastive Learning with Node Similarity

Subgraph Networks Based Contrastive Learning

GCL: Contrastive learning instead of graph convolution for node classification

Tensor-Fused Multi-View Graph Contrastive Learning

Semi-Supervised and Self-Supervised Classification with Multi-View Graph Neural Networks

Adaptive Multi-layer Contrastive Graph Neural Networks

Localized Graph Contrastive Learning

Enhancing Graph Contrastive Learning with Reliable and Informative Augmentation for Recommendation