Research map / Graph learning / Graph self-supervised learning
Khan-GCL: Kolmogorov–Arnold Network Based Graph Contrastive Learning with Hard Negatives
Zihu Wang, Boxun Xu, Hejia Geng, Peng Li
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace MLP encoder with KAN encoder in GraphCL (no hard negatives) (ablation) | GraphCL · MUTAG | accuracy: 86.8 -> 88.4 | Improved |
| replace MLP encoder with KAN encoder in GraphCL (no hard negatives) (ablation) | GraphCL · TU-datasets average | average accuracy: 75.7 -> 76.8 | Improved |
| replace MLP encoder with KAN encoder in JOAOv2 (ablation) | JOAOv2 · MUTAG | accuracy: 87.7 -> 88.6 | Improved |
| add CKFI hard negatives on top of KAN encoder (ablation) | Khan-GCL · TU-datasets average | average accuracy: 76.8 -> 78.6 | Improved |
| add Khan-GCL hard negatives to JOAOv2 with KAN (ablation) | JOAOv2 · MUTAG | accuracy: 88.6 -> 92.1 | Improved |
| use random Gaussian perturbation on all dimensions instead of CKFI-guided perturbation (ablation) | Khan-GCL · TU-datasets average | average accuracy: 76.8 -> 76.6 | No clear effect |
| add discriminative dimension perturbation to independent-only perturbation (ablation) | Khan-GCL · TU-datasets average | average accuracy: 77.7 -> 78.6 | Improved |
| add independent dimension perturbation to discriminative-only perturbation (ablation) | Khan-GCL · TU-datasets average | average accuracy: 77.8 -> 78.6 | Improved |
| use both CKFI feature types in perturbation (ablation) | Khan-GCL · TU-datasets | accuracy | Improved |
| Khan-GCL vs CI-GCL on some individual datasets | Khan-GCL · BBBP, BACE transfer | ROC-AUC: BBBP 73.5 vs 74.4; BACE 80.9 vs 84.4 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- TopoGCL — TopoGCL: Topological Graph Contrastive Learning (AAAI 2024) · better on 0 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Does GCL Need a Large Number of Negative Samples? Enhancing Graph Contrastive Learning with Effective and Efficient Negative Sampling (AAAI 2025 · similarity 0.94)
- ReGCL: Rethinking Message Passing in Graph Contrastive Learning (AAAI 2024 · similarity 0.91)
- S3GCL: Spectral, Swift, Spatial Graph Contrastive Learning (ICML 2024 · similarity 0.91)
- Hierarchical Topology Isomorphism Expertise Embedded Graph Contrastive Learning (AAAI 2024 · similarity 0.90)
- A New Mechanism for Eliminating Implicit Conflict in Graph Contrastive Learning (AAAI 2024 · similarity 0.90)
- Beyond Homophily: Graph Contrastive Learning with Macro-Micro Message Passing (AAAI 2025 · similarity 0.90)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
