Research map / Graph learning / Graph self-supervised learning
Contrastive Laplacian Eigenmaps
Hao Zhu, Ke Sun, Piotr Koniusz
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 |
|---|---|---|---|
| use COLES-S²GC instead of S²GC | S²GC · Cora (5 labeled samples/class) | accuracy (%): 71.4±4.4 -> 76.5±2.6 | Improved |
| use COLES-S²GC instead of S²GC | S²GC · Citeseer (5 labeled samples/class) | accuracy (%): 60.3±4.0 -> 67.5±2.2 | Improved |
| use COLES-S²GC instead of S²GC | S²GC · Cora Full (5 labeled samples/class) | accuracy (%): 41.8±1.7 -> 50.8±1.4 | Improved |
| use COLES-S²GC instead of S²GC | S²GC · Ogbn-arxiv | test accuracy (%): 72.01±0.25 -> 72.48±0.25 | Improved |
| remove negative Laplacian graphs (κ=0) (ablation) | COLES-S²GC · Cora (5 labeled samples/class) | accuracy (%): 76.24 -> 70.12 | Hurt |
| add more than one negative Laplacian graph (ablation) | COLES-S²GC · Cora (5 labeled samples/class) | accuracy (%): 76.24 -> 75.89 -> 75.79 | No clear effect |
| remove negative Laplacian graphs (κ=0) (ablation) | COLES-GCN · Citeseer (5 labeled samples/class) | accuracy (%): 65.85 -> 45.31 | Hurt |
| replace the geometric mean with the arithmetic mean in the uniformity loss (ablation) | COLES-S²GC · Cora (5 labeled samples/class) | accuracy (%): 76.5±2.6 -> 71.8±3.0 | Hurt |
| replace the geometric mean with the harmonic mean in the uniformity loss (ablation) | COLES-S²GC · Cora (5 labeled samples/class) | accuracy (%): 76.5±2.6 -> 75.2±3.5 | Hurt |
| replace the geometric mean with the quadratic mean in the uniformity loss (ablation) | COLES-S²GC · Cora (5 labeled samples/class) | accuracy (%): 76.5±2.6 -> 72.3±2.5 | Hurt |
| use Stiefel optimization instead of soft orthogonality (ablation) | COLES-GCN · Cora (5 labeled samples/class) | accuracy (%): 73.8±3.4 -> 75.0±3.4 | Improved |
| add data augmentation (ablation) | COLES-GCN · Cora (5 labeled samples/class) | accuracy (%): 73.8±3.4 -> 75.3±3.3 | Improved |
Closest papers by meaning
- S3GCL: Spectral, Swift, Spatial 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 0.88)
- ReGCL: Rethinking Message Passing in Graph Contrastive Learning (AAAI 2024 · similarity 0.88)
- Single-View Graph Contrastive Learning with Soft Neighborhood Awareness (AAAI 2025 · similarity 0.88)
- A New Mechanism for Eliminating Implicit Conflict in Graph Contrastive Learning (AAAI 2024 · similarity 0.87)
- Efficient Contrastive Learning for Fast and Accurate Inference on Graphs (ICML 2024 · similarity 0.87)
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.
