Research map / Graph learning / Graph self-supervised learning
GCL-OT: Graph Contrastive Learning with Optimal Transport for Heterophilic Text-Attributed Graphs
Yating Ren, Yikun Ban, Huobin Tan
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 InfoNCE with the GCL-OT loss (ablation) | GCL-OT · Cora with perturbed edges and texts | accuracy | Improved |
| add the GCL-OT contrastive loss (ablation) | GCL-OT · Cora and other datasets | accuracy | Improved |
| add the alignment loss L_MHA (RealSoftMax and filter prompt) (ablation) | GCL-OT · Texas and Cornell | accuracy | Improved |
| add the latent homophily mining loss L_LHM (ablation) | GCL-OT · Amazon | accuracy | Improved |
| swap DistilBERT for stronger PLMs (DeBERTa, RoBERTa) (ablation) | GCL-OT · Wisconsin, Cornell, Texas | accuracy | No clear effect |
| add GCL-OT to GCN and GAT under edge deletion (-500 edges) | GCN, GAT · Cora edge perturbation | accuracy: relative gains of 24.74% and 79.44% | Improved |
| use GCL-OT with DistilBERT under text perturbation | DistilBERT · Cora text perturbation | accuracy: retains nearly 85% | Improved |
| vary lambda, beta and LRSinkhorn iterations (ablation) | GCL-OT (GCN) · Cora | accuracy | Mixed |
| use unsupervised GCL-OT versus heterophilic GCL baselines | GCL-OT (GCN) · Wisconsin unsupervised | accuracy (%): 71.93 -> 72.83 | Improved |
Closest papers by meaning
- ReGCL: Rethinking Message Passing in Graph Contrastive Learning (AAAI 2024 · similarity 0.90)
- Contrastive Auxiliary Learning with Structure Transformation for Heterogeneous Graphs (AAAI 2025 · similarity 0.89)
- Beyond Homophily: Graph Contrastive Learning with Macro-Micro Message Passing (AAAI 2025 · similarity 0.89)
- Perfect Alignment May be Poisonous to Graph Contrastive Learning (ICML 2024 · similarity 0.89)
- Does GCL Need a Large Number of Negative Samples? Enhancing Graph Contrastive Learning with Effective and Efficient Negative Sampling (AAAI 2025 · similarity 0.89)
- Graph Structure Refinement with Energy-based Contrastive Learning (AAAI 2025 · similarity 0.88)
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.
