Research map / Graph learning / GNN architectures
Equivariant Hypergraph Neural Networks
Jinwoo Kim, Saeyoon Oh, Sungjun Cho, Seunghoon Hong
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 global interaction and order embedding (vs removing both) (ablation) | EHNN-MLP · k-edge identification (seen k, ablation) | accuracy (%): 78.94 -> 98.02 | Improved |
| add order embedding (vs removing it) (ablation) | EHNN-MLP · k-edge identification (seen k, ablation) | accuracy (%): 84.05 -> 98.02 | Improved |
| add global interaction (vs removing it) (ablation) | EHNN-MLP · k-edge identification (seen k, ablation) | accuracy (%): 80.34 -> 98.02 | Improved |
| use MLP realization instead of naive lookup table (ablation) | EHNN · k-edge identification (seen k, ablation) | accuracy (%): 87.09 -> 98.02 | Improved |
| use MLP realization instead of naive hypernetwork (ablation) | EHNN · k-edge identification (seen k, ablation) | accuracy (%): 83.74 -> 98.02 | Improved |
| replace message passing AllDeepSets with EHNN-MLP (ablation) | AllDeepSets · k-edge identification (seen k) | accuracy (%): 76.99 -> 98.02 | Improved |
| replace AllDeepSets with EHNN-MLP on unseen orders (extrapolation) | AllDeepSets · k-edge identification (unseen k, extrapolation) | accuracy (%): 79.01 -> 85.65 | Improved |
| add attention (EHNN-Transformer vs EHNN-MLP) on seen k (ablation) | EHNN-MLP · k-edge identification (seen k) | accuracy (%): 98.02 -> 99.69 | Improved |
| add attention (EHNN-Transformer vs EHNN-MLP) on extrapolation (ablation) | EHNN-MLP · k-edge identification (unseen k, extrapolation) | accuracy (%): 85.65 -> 90.19 | Improved |
| add attention (EHNN-Transformer vs EHNN-MLP) on Walmart(1) | EHNN-MLP · Walmart(1) semi-supervised | accuracy (%): 65.65 -> 68.73 | Improved |
| add attention (EHNN-Transformer vs EHNN-MLP) on Willow | EHNN-MLP · Willow keypoint matching (avg) | accuracy (%): 96.33 -> 97.08 | Improved |
| add attention (EHNN-Transformer vs EHNN-MLP) on PASCAL-VOC | EHNN-MLP · PASCAL-VOC keypoint matching (avg) | accuracy (%): 79.90 -> 79.74 | No clear effect |
Closest papers by meaning
- Are High-Degree Representations Really Unnecessary in Equivariant Graph Neural Networks? (NeurIPS 2024 · similarity 0.88)
- Training-Free Message Passing for Learning on Hypergraphs (ICLR 2025 · similarity 0.88)
- Universally Invariant Learning in Equivariant GNNs (NeurIPS 2025 · similarity 0.87)
- On Explaining Equivariant Graph Networks via Improved Relevance Propagation (ICML 2025 · similarity 0.87)
- Deep Hypergraph Neural Networks with Tight Framelets (AAAI 2025 · similarity 0.87)
- When Hypergraph Meets Heterophily: New Benchmark Datasets and Baseline (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.
