Mitigating Local Cohesion and Global Sparseness in Graph Contrastive Learning with Fuzzy Boundaries

Graph contrastive learning (GCL) aims at narrowing positives while dispersing negatives, often causing a minority of samples with great similarities to gather as a small group. It results in two latent shortcomings in GCL: 1) local cohesion that a class cluster contains numerous independent small groups, and 2) global sparseness that these small groups (or isolated samples) dispersedly distribute among all clusters. These shortcomings make the learned distribution only focus on local similarities among partial samples, which hinders the ability to capture the ideal global structural properties among real clusters, especially high intra-cluster compactness and inter-cluster separateness. Considering this, we design a novel fuzzy boundary by extending the original cluster boundary with fuzzy set theory, which involves fuzzy boundary construction and fuzzy boundary contraction to address these shortcomings. The fuzzy boundary construction dilates the original boundaries to bridge the local groups, and the fuzzy boundary contraction forces the dispersed samples or groups within the fuzzy boundary to gather tightly, jointly mitigating local cohesion and global sparseness while forming the ideal global structural distribution. Extensive experiments demonstrate that a graph auto-encoder with the fuzzy boundary significantly outperforms current state-of-the-art GCL models in both downstream tasks and quantitative analysis.

Many AI models learn by looking at how things are connected — like who is friends with whom, or which products are often bought together. One popular method, called graph contrastive learning, helps the model learn by pulling similar samples closer and pushing different ones apart.However, we noticed a problem: when samples are very similar, the model tends to group them into small, tight clusters that end up scattered and far from others. This makes it hard for the model to understand the full picture of how different groups are organized.To fix this, we introduced a new technique that gently expands the boundaries around each group. This allows nearby small clusters to be included together. Then, we bring these items closer within the group, so each group is clearer and more complete. This simple idea helps the model organize information better and leads to improved results in tasks like classification. Our method makes AI models more reliable and better at learning from graph-based data.