Research map / Graph learning / Graph self-supervised learning
Revisiting Graph Contrastive Learning from the Perspective of Graph Spectrum
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 |
|---|---|---|---|
| Add SpCo to CCA-SSG | CCA-SSG · Cora | Ma-F1: 82.9±0.8 -> 83.6±0.4 | Helped |
| Add SpCo to CCA-SSG | CCA-SSG · Citeseer | Ma-F1: 67.9±1.0 -> 68.5±1.0 | Helped |
| Add SpCo to CCA-SSG | CCA-SSG · BlogCatalog | Ma-F1: 72.0±0.5 -> 72.8±0.3 | Helped |
| Add SpCo to CCA-SSG | CCA-SSG · Flickr | Ma-F1: 37.0±1.1 -> 38.7±0.6 | Helped |
| Add SpCo to CCA-SSG | CCA-SSG · PubMed | Ma-F1: 80.7±0.6 -> 81.3±0.3 | Helped |
| Add SpCo to GRACE | GRACE · Citeseer | Ma-F1: 65.1±1.2 -> 65.1±0.8 | No clear effect |
| Add SpCo to DGI | DGI · Flickr | Ma-F1: 31.2±1.6 -> 33.7±0.7 | Helped |
| Use PPR matrix instead of the other tested augmentations | Case-study GCL · Cora | accuracy: 34.9$\pm3.5$ to 58.0$\pm1.6$ across tested augmentations | Helped |
| Retain the lowest-frequency components in the generated view | Case-study GCL · Cora, Citeseer, BlogCatalog, Flickr | accuracy | Helped |
| Include more high-frequency components in the generated view | Case-study GCL · Cora, Citeseer, BlogCatalog, Flickr | accuracy | Helped |
| Use Laplacian as the spectral guidance matrix rather than I+L or I+L+L² (ablation) | SpCo · Cora and BlogCatalog | Micro-F1 | Helped |
| Set entropy regularization strength too small (ablation) | SpCo · BlogCatalog | Micro-F1 | Hurt |
Closest papers by meaning
- Community-Invariant Graph Contrastive Learning (ICML 2024 · similarity 0.93)
- Perfect Alignment May be Poisonous to Graph Contrastive Learning (ICML 2024 · similarity 0.93)
- Graph Contrastive Invariant Learning from the Causal Perspective (AAAI 2024 · similarity 0.93)
- Graph Contrastive Learning with Joint Spectral Augmentation of Attribute and Topology (AAAI 2025 · similarity 0.93)
- S3GCL: Spectral, Swift, Spatial Graph Contrastive Learning (ICML 2024 · similarity 0.92)
- ReGCL: Rethinking Message Passing in Graph Contrastive Learning (AAAI 2024 · similarity 0.92)
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.