<ul><li>Graph contrastive learning (GCL) has shown potential for learning graph representations from unlabeled data.</li><li>Khan-GCL addresses limitations in conventional GCL methods by integrating Kolmogorov-Arnold Network (KAN) for enhanced representational capacity.</li><li>Khan-GCL introduces novel techniques for generating semantically meaningful hard negative samples, improving graph representation learning.</li><li>Extensive experiments confirm Khan-GCL's state-of-the-art performance compared to other GCL methods on various datasets and tasks.</li></ul>

Khan-GCL: Kolmogorov-Arnold Network Based Graph Contrastive Learning with Hard Negatives

Discover more