Research map / Graph learning / Graph self-supervised learning
Neighbor Contrastive Learning on Learnable Graph Augmentation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| use learnable attention-based graph augmentation and neighbor contrastive learning | NCLA · Cora, 1 label/class | accuracy: 59.1±10.9 -> 63.1±11.2 | Helped |
| use learnable attention-based graph augmentation and neighbor contrastive learning | NCLA · Citeseer, 1 label/class | accuracy: 48.6±11.3 -> 52.2±13.5 | Helped |
| use learnable attention-based graph augmentation and neighbor contrastive learning | NCLA · PubMed, 1 label/class | accuracy: 57.7±10.5 -> 60.2±12.4 | Helped |
| use learnable attention-based graph augmentation and neighbor contrastive learning | NCLA · Coauthor-CS, 1 label/class | accuracy: 79.8±8.0 -> 83.0±6.2 | Helped |
| use learnable attention-based graph augmentation and neighbor contrastive learning | NCLA · Amazon-Photo, 1 label/class | accuracy: 71.6±6.2 -> 75.6±6.0 | Helped |
| replace neighbor contrastive loss with InfoNCE (ablation) | NCLA · Cora, Citeseer, PubMed, Coauthor-CS, Amazon-Photo | accuracy: 63.1 -> 60.9; 52.2 -> 48.5; 60.2 -> 59.7; 83.0 -> 80.4; 75.6 -> 75.3 | Hurt |
| replace neighbor contrastive loss with NT-Xent (ablation) | NCLA · Cora, Citeseer, PubMed, Coauthor-CS, Amazon-Photo | accuracy: 63.1 -> 60.1; 52.2 -> 51.6; 60.2 -> 60.0; 83.0 -> 80.5; 75.6 -> 75.2 | Hurt |
| change intra-view neighbors from positives to negatives (ablation) | NCLA · Cora, Citeseer, PubMed, Coauthor-CS, Amazon-Photo | accuracy: 63.1 -> 62.4; 52.2 -> 51.0; 60.2 -> 59.6; 83.0 -> 82.7; 75.6 -> 72.5 | Hurt |
| change inter-view neighbors from positives to negatives (ablation) | NCLA · Cora, Citeseer, PubMed, Coauthor-CS, Amazon-Photo | accuracy: 63.1 -> 62.3; 52.2 -> 49.5; 60.2 -> 59.4; 83.0 -> 81.4; 75.6 -> 71.2 | Hurt |
| increase training labels from 1 to 20 per class | NCLA · Cora, Citeseer, PubMed, Coauthor-CS, Amazon-Photo | accuracy: 63.1±11.2 -> 82.2±1.6; 52.2±13.5 -> 71.7±0.9; 60.2±12.4 -> 82.0±1.4; 83.0±6.2 -> 91.5±0.7; 75.6±6.0 -> 90.2±1.3 | Helped |
Closest papers by meaning
- Community-Invariant Graph Contrastive Learning (ICML 2024 · similarity 0.93)
- Hierarchical Topology Isomorphism Expertise Embedded Graph Contrastive Learning (AAAI 2024 · similarity 0.92)
- Edge Contrastive Learning: An Augmentation-Free Graph Contrastive Learning Model (AAAI 2025 · similarity 0.92)
- Graph Contrastive Learning with Joint Spectral Augmentation of Attribute and Topology (AAAI 2025 · similarity 0.91)
- Does GCL Need a Large Number of Negative Samples? Enhancing Graph Contrastive Learning with Effective and Efficient Negative Sampling (AAAI 2025 · similarity 0.91)
- S3GCL: Spectral, Swift, Spatial Graph Contrastive Learning (ICML 2024 · similarity 0.91)
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.