Research map / Graph learning / GNN architectures
Sheaf Hypergraph Networks
Iulia Duta, Giulia Cassarà, Fabrizio Silvestri, Pietro Lio
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 |
|---|---|---|---|
| replace trivial-sheaf Laplacian with learnable linear sheaf Laplacian (HyperGNN -> SheafHyperGNN) | HyperGNN · Senate | accuracy (%): 48.59 -> 68.73 | Improved |
| replace trivial-sheaf with sheaf in non-linear Laplacian (HyperGCN -> SheafHyperGCN) | HyperGCN · Senate | accuracy (%): 51.13 -> 66.33 | Improved |
| replace HyperGNN with SheafHyperGNN | HyperGNN · House | accuracy (%): 61.39 -> 73.84 | Improved |
| replace HyperGNN with SheafHyperGNN | HyperGNN · Cora | accuracy (%): 79.39 -> 81.30 | Improved |
| replace HyperGNN with SheafHyperGNN | HyperGNN · synthetic heterophilic alpha=7 | accuracy (%): 63.8 -> 77.3 | Improved |
| use linear sheaf Laplacian (SheafHyperGNN) instead of non-linear (SheafHyperGCN) | SheafHyperGNN · Cora_CA | accuracy (%): 83.26 -> 85.52 | Improved |
| use general d x d restriction maps instead of diagonal (ablation) | SheafHyperGNN · Cora_CA | accuracy (%): 85.52 -> 77.12 | Hurt |
| use low-rank restriction maps instead of diagonal (ablation) | SheafHyperGNN · Congress | accuracy (%): 91.81 -> 74.83 | Hurt |
| use low-rank restriction maps instead of diagonal (ablation) | SheafHyperGCN · Congress | accuracy (%): 90.37 -> 84.88 | Hurt |
| use general restriction maps instead of diagonal (ablation) | SheafHyperGCN · Senate | accuracy (%): 66.33 -> 65.49 | No clear effect |
| use low-rank restriction maps instead of diagonal (ablation) | SheafHyperGNN · Senate | accuracy (%): 68.73 -> 68.45 | No clear effect |
| increase stalk dimension d from 1 to >1 (ablation) | SheafHyperGNN · synthetic heterophilic alpha=7 | accuracy | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- HyperGCN — HyperGCN: A New Method For Training Graph Convolutional Networks on Hypergraphs (NeurIPS 2019) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Directional Sheaf Hypergraph Networks: Unifying Learning on Directed and Undirected Hypergraphs (ICLR 2026 · similarity 0.95)
- High-Pass Matters: Theoretical Insights and Sheaflet-Based Design for Hypergraph Neural Networks (AAAI 2026 · similarity 0.92)
- Higher-Order Learning with Graph Neural Networks via Hypergraph Encodings (NeurIPS 2025 · similarity 0.90)
- SPHINX: Structural Prediction using Hypergraph Inference Network (ICML 2025 · similarity 0.88)
- Mitigating Over-squashing without Rewiring: A Sheaf Effective Resistance Perspective (NeurIPS 2026 · similarity 0.88)
- Training-Free Message Passing for Learning on Hypergraphs (ICLR 2025 · similarity 0.88)
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.
