Research map / Graph learning / Graph self-supervised learning
Augmentations in Hypergraph Contrastive Learning: Fabricated and Generative
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 generalized hyperedge perturbation (A2) contrastive learning | SetGNN · Cora 10% labels | accuracy: 67.93 -> 72.58 | Helped |
| use generative augmentation A6 instead of best fabricated A2 | HyperGCL · Yelp 10% labels | accuracy: 31.39 -> 34.64 | Helped |
| naive hyperedge perturbation A1 instead of A2 (ablation) | HyperGCL · Cora 10% labels | accuracy: 72.58 -> 72.39 | Hurt |
| replace multi-task learning with pretrain + linear evaluation (ablation) | HyperGCL A2 · Cora 10% labels | accuracy: 72.58 -> 52.59 | Hurt |
| replace multi-task learning with pretrain + finetune (ablation) | HyperGCL A2 · Cora 10% labels | accuracy: 72.58 -> 68.39 | Hurt |
| contrast on clique-expanded graph with HGNN instead of hypergraph SetGNN (ablation) | HyperGCL · Cora 10% labels | accuracy: 72.58 -> 67.18 | Hurt |
| subgraph augmentation (A5) added to SetGNN (ablation) | SetGNN · Cora 10% labels | accuracy: 67.93 -> 67.96 | No clear effect |
| attribute masking vs generative augmentation at 1% labels (ablation) | HyperGCL · Cora 1% labels | accuracy: 57.45 vs 58.55 | Mixed |
| add generative augmentation A6 under Minmax attack | SetGNN · Cora robustness | accuracy: 66.58 -> 71.98 | Helped |
| use A6 on fairness data | SetGNN · German Credit | Delta_EO: 4.06 -> 0.47 | Helped |
Closest papers by meaning
- HyperAim: Hypergraph Contrastive Learning with Adaptive Multi-frequency Filters (AAAI 2026 · similarity 0.90)
- HypeBoy: Generative Self-Supervised Representation Learning on Hypergraphs (ICLR 2024 · similarity 0.89)
- When Hypergraph Meets Heterophily: New Benchmark Datasets and Baseline (AAAI 2025 · similarity 0.88)
- Perfect Alignment May be Poisonous to Graph Contrastive Learning (ICML 2024 · similarity 0.88)
- Community-Invariant Graph Contrastive Learning (ICML 2024 · similarity 0.88)
- Adversarial Contrastive Graph Augmentation with Counterfactual Regularization (AAAI 2025 · 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.