Research map / Graph learning / Graph self-supervised learning
CCAHCL: Multi-Level Hypergraph Contrastive Learning for Connected Component Awareness
Zhuo Li, Gengyu Lyu, Yuena Lin, Ziang Chen, Zhiyuan Ma, Zhen Yang, Zun Li
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 |
|---|---|---|---|
| add hyperedge-level loss to node-level loss (ablation) | CCAHCL · Cora-C | accuracy (%): 82.10 -> 83.08 | Improved |
| add connected-component-level loss to node+hyperedge losses (ablation) | CCAHCL · Cora-C | accuracy (%): 83.08 -> 83.51 | Improved |
| add connected-component-level loss to node+hyperedge losses (ablation) | CCAHCL · Citeseer | accuracy (%): 73.60 -> 73.84 | Improved |
| add connected-component-level loss to node+hyperedge losses (ablation) | CCAHCL · ModelNet40 | accuracy (%): 97.57 -> 97.63 | Improved |
| add connected-component-level loss to node loss only (ablation) | CCAHCL · Cora-C | accuracy (%): 82.10 -> 82.36 | Improved |
| use only connected-component-level loss instead of node-level loss (ablation) | CCAHCL · Cora-C | accuracy (%): 82.10 -> 64.40 | Hurt |
| use only hyperedge-level loss instead of node-level loss (ablation) | CCAHCL · Cora-A | accuracy (%): 83.70 -> 77.46 | Hurt |
| full CCAHCL vs TriCL | CCAHCL · Cora-C | accuracy (%): 81.57 -> 83.51 | Improved |
| full CCAHCL vs TriCL | CCAHCL · Citeseer | accuracy (%): 72.02 -> 73.84 | Improved |
| full CCAHCL vs TriCL | CCAHCL · DBLP | accuracy (%): 91.12 -> 91.40 | Improved |
| full CCAHCL vs TriCL in k-means clustering | CCAHCL · Citeseer | pairwise F1: 57.4 -> 56.7 | No clear effect |
Closest papers by meaning
- I’m Me, We’re Us, and I’m Us: Tri-directional Contrastive Learning on Hypergraphs (AAAI 2023 · similarity 0.90)
- H$^2$CL: Heterogeneity-Aware Hypergraph Contrastive Learning for Robust Representation Learning (ICML 2026 · similarity 0.90)
- HyperNoRA: Hyperedge Prediction via Node-Level Relation-Aware Self-Supervised Hypergraph Learning (AAAI 2026 · similarity 0.90)
- Hypergraph Representation Learning with Hyperlink Random Effects (NeurIPS 2026 · similarity 0.87)
- ReGCL: Rethinking Message Passing in Graph Contrastive Learning (AAAI 2024 · similarity 0.86)
- Does GCL Need a Large Number of Negative Samples? Enhancing Graph Contrastive Learning with Effective and Efficient Negative Sampling (AAAI 2025 · similarity 0.86)
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.
