Research map / Graph learning / Graph self-supervised learning
Spectral Augmentation for Self-Supervised Learning on Graphs
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 uniform topology augmentation with spectral-change-guided augmentation | GCL on node-classification benchmarks · Cora, PubMed, Wiki-CS, Amazon-Photo, Coauthor-CS | accuracy (%): 85.16$\pm$0.52 -> 85.86$\pm$0.57; 80.13$\pm$0.84 -> 81.54$\pm$0.24; 80.14$\pm$0.48 -> 82.13$\pm$0.15; 93.17$\pm$0.30 -> | Helped |
| Use spectral augmentation instead of uniform topology augmentation | GCL on Amazon-Computer · Amazon-Computer | accuracy (%): 90.34$\pm$0.19 -> 90.09$\pm$0.32 | Hurt |
| Use spectral-guided augmentation rather than the strongest reported baseline | GCL graph classification · PROTEINS, DD, COLLAB, REDDIT-MULTI-5K, IMDB-BINARY, IMDB-MULTI | accuracy (%): 74.02$\pm$0.32 -> 75.78$\pm$0.41; 75.23$\pm$0.39 -> 75.78$\pm$0.52; 73.32$\pm$0.61 -> 75.01$\pm$0.45; 53.05$\pm$0.40 -> | Helped |
| Use spectral-guided augmentation rather than the strongest reported baseline | GCL graph classification · NCI1, MUTAG, REDDIT-BINARY | accuracy (%): 72.99$\pm$0.75 -> 71.43$\pm$0.49; 89.25$\pm$1.45 -> 89.12$\pm$0.76; 85.52$\pm$0.79 -> 83.62$\pm$0.64 | Hurt |
| Use spectral-guided augmentation rather than the strongest reported baseline | GCL graph regression · mollipo, molfreesolv | RMSE: 0.842$\pm$0.028 -> 0.802$\pm$0.019; 5.131$\pm$0.722 -> 4.531$\pm$0.463 | Helped |
| Use spectral-guided augmentation rather than the strongest reported baseline | GCL graph regression · molesol | RMSE: 1.217$\pm$0.087 -> 1.218$\pm$0.052 | No clear effect |
| Use spectral-guided augmentation rather than uniform augmentation | GCL transfer pretraining · Tox21, SIDER, BACE, ToxCast, PPI | ROC-AUC (%): 76.5$\pm$0.8 -> 78.0$\pm$0.5; 63.3$\pm$0.8 -> 64.7$\pm$0.5; 78.5$\pm$0.8 -> 79.9$\pm$0.7; 63.2$\pm$0.5 -> 64.2$\pm$0.4; | Helped |
| Use spectral-guided augmentation rather than the strongest baseline | GCL transfer pretraining · ClinTox | ROC-AUC (%): 81.0$\pm$1.6 -> 80.7$\pm$2.1 | Hurt |
| Use spectral-guided augmentation rather than the strongest baseline | GCL transfer pretraining · BBBP, HIV, MUV | ROC-AUC (%): 71.4$\pm$0.9 -> 70.0$\pm$0.7; 78.5$\pm$1.2 -> 77.8$\pm$0.6; 75.3$\pm$2.5 -> 73.8$\pm$0.9 | Mixed |
| Use spectral-guided augmentation rather than uniform augmentation | GCL under adversarially poisoned Cora graphs · Cora, Mettack ratio 0.2 | accuracy (%): 63.93$\pm$0.58 -> 69.92$\pm$0.83 | Helped |
| Allocate the same edge-removal budget toward inter-cluster edges rather than uniformly | GRACE on Cora · Cora | F1 score | Helped |
Closest papers by meaning
- Community-Invariant Graph Contrastive Learning (ICML 2024 · similarity 0.93)
- Graph Contrastive Learning with Joint Spectral Augmentation of Attribute and Topology (AAAI 2025 · similarity 0.91)
- Hierarchical Topology Isomorphism Expertise Embedded Graph Contrastive Learning (AAAI 2024 · similarity 0.90)
- Graph Contrastive Invariant Learning from the Causal Perspective (AAAI 2024 · similarity 0.90)
- Through the Dual-Prism: A Spectral Perspective on Graph Data Augmentation for Graph Classifications (AAAI 2025 · similarity 0.90)
- S3GCL: Spectral, Swift, Spatial Graph Contrastive Learning (ICML 2024 · 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.