Research map / Graph learning / GNN architectures
From Hypergraph Energy Functions to Hypergraph Neural Networks
Yuxin Wang, Quan Gan, Xipeng Qiu, Xuanjing Huang, David Wipf
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 both clique and star expansion terms together (vs clique-only or star-only) (ablation) | PhenomNN_simple · Cora (co-authorship), Pubmed (co-citation) | accuracy: Cora: 77.62 full vs 77.06 clique-only / 77.28 star-only; Pubmed: 76.20 star-only ties full, clique-only 75.71 | Mixed |
| use both clique and star expansion terms together (vs clique-only or star-only) (ablation) | PhenomNN · Cora (co-authorship), Pubmed (co-citation), Citeseer | accuracy: Cora 77.11 vs 76.74 clique-only / 76.83 star-only; Pubmed 78.12 vs 77.94 / 77.52 | Improved |
| add ReLU-based proximal energy with simplified energy (psi={W,I,I}) instead of full parameterized energy | PhenomNN_simple vs PhenomNN · Cora (co-authorship) and co-citation datasets | accuracy: PhenomNN_simple higher on Cora (77.62 vs 77.11) and Cora-cocitation tie; PhenomNN higher on DBLP, Pubmed, Citeseer, NTU2 | Mixed |
| hidden dimension reduced from 512 to 64 in PhenomNN_simple (ablation) | PhenomNN_simple · House(1) and House(0.6) | accuracy: House(1) 71.77 at 512 vs 68.66 at 64; House(0.6) 86.96 vs 85.44 | Hurt |
| hidden dimension reduced from 512 to 64 in PhenomNN_simple (ablation) | PhenomNN_simple · ModelNet40* | accuracy: 98.66 at 512 vs 98.51 at 64 | No clear effect |
| PhenomNN (energy-unfolded hypergraph layers) vs strongest baseline on co-citation/co-authorship benchmarks | PhenomNN · Pubmed (co-citation), Citeseer, DBLP | accuracy: Pubmed 78.12 vs 75.8 (H-GCNII); Citeseer 65.77 vs 64.5 | Improved |
| PhenomNN_simple vs AllSetTransformer on visual/ModelNet40* and NTU2012* benchmarks | PhenomNN_simple · NTU2012*, ModelNet40*, House(1) | accuracy: NTU2012* 91.03 vs 88.69; ModelNet40* 98.66 vs 98.20; House(1) 71.77 vs 69.33 | Improved |
| PhenomNN_simple vs AllSetTransformer on Yelp | PhenomNN_simple · Yelp | accuracy: 32.26 vs 36.89 | Hurt |
Closest papers by meaning
- Training-Free Message Passing for Learning on Hypergraphs (ICLR 2025 · similarity 0.90)
- Hypergraph Neural Architecture Search (AAAI 2024 · similarity 0.89)
- When Hypergraph Meets Heterophily: New Benchmark Datasets and Baseline (AAAI 2025 · similarity 0.88)
- Improved Expressivity of Hypergraph Neural Networks through High-Dimensional Generalized Weisfeiler-Leman Algorithms (ICML 2025 · similarity 0.88)
- Hypergraph-enhanced Dual Semi-supervised Graph Classification (ICML 2024 · similarity 0.88)
- LightHGNN: Distilling Hypergraph Neural Networks into MLPs for 100x Faster Inference (ICLR 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.
