Research map / Graph learning / Graph self-supervised learning
Rethinking and Scaling Up Graph Contrastive Learning: An Extremely Efficient Approach with Group Discrimination
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 |
|---|---|---|---|
| replace DGI's bilinear discriminator and learned summary with simple BCE group discrimination (DGI_BCE) (ablation) | DGI · Cora | accuracy (%): 81.7 -> 82.5 | Helped |
| DGI_BCE vs DGI memory (ablation) | DGI · PubMed | memory (MB): 11471MB -> 1629MB | Helped |
| DGI_BCE vs DGI training time per epoch (ablation) | DGI · Cora | training time per epoch (s): 0.085s -> 0.010s | Helped |
| set constant summary vector to nonzero value (0.2 to 1.0) instead of 0 (ablation) | DGI · Cora | accuracy (%): 70.3 -> 82.4 | Helped |
| vary constant summary vector value among 0.2 to 1.0 (ablation) | DGI · Cora, CiteSeer, PubMed | accuracy (%): 82.4 / 82.3 / 82.5 on Cora | No clear effect |
| GGD (group discrimination, 1 epoch) vs GBT (300 epochs) on ogbn-arxiv total training time | GGD · ogbn-arxiv | total training time (s): 1,941.00 -> 0.18 | Helped |
| GGD (hidden 256) vs GBT on ogbn-arxiv accuracy | GGD · ogbn-arxiv test | accuracy (%): 70.1 -> 70.3 | Helped |
| GGD vs GBT on ogbn-products | GGD · ogbn-products test | accuracy (%): 70.5 -> 75.7 | Helped |
| GGD vs BGRL/GBT on ogbn-papers100M | GGD · ogbn-papers100M test | accuracy (%): 61.5 -> 63.5 (GBT) | Mixed |
| GGD vs best baseline on CiteSeer | GGD · CiteSeer | accuracy (%): 73.6 (CG3) -> 73.0 | Hurt |
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 0.92)
- S3GCL: Spectral, Swift, Spatial Graph Contrastive Learning (ICML 2024 · similarity 0.91)
- Exploitation of a Latent Mechanism in Graph Contrastive Learning: Representation Scattering (NeurIPS 2024 · similarity 0.90)
- Hierarchical Topology Isomorphism Expertise Embedded Graph Contrastive Learning (AAAI 2024 · similarity 0.90)
- ReGCL: Rethinking Message Passing in Graph Contrastive Learning (AAAI 2024 · similarity 0.90)
- Improving Graph Contrastive Learning via Adaptive Positive Sampling (CVPR 2024 · similarity 0.89)
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.