Research map / Graph learning / GNN architectures
Directional Sheaf Hypergraph Networks: Unifying Learning on Directed and Undirected Hypergraphs
Emanuele Mule, Stefano Fiorini, Antonio Purificato, Federico Siciliano, Stefano Coniglio, Fabrizio Silvestri
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 |
|---|---|---|---|
| Use the directed sheaf hypergraph Laplacian (DSHNLight) instead of the best baseline | DSHNLight · email-EU | accuracy: 68.91 -> 82.67 | Improved |
| Use DSHNLight instead of GeDi-HNN on Roman-empire | DSHNLight · Roman-empire | accuracy: 83.87 -> 89.24 | Improved |
| Add directional sheaf Laplacian versus SheafHyperGNN on Telegram | DSHN · Telegram | accuracy: 70.00 -> 88.65 | Improved |
| Use DSHN on Cora versus PhenomNN | DSHN · Cora | accuracy: 88.12 -> 88.02 | No clear effect |
| Use nonzero charge parameter q (directional information) on Cora (ablation) | DSHN · Cora | accuracy | Hurt |
| Use nonzero charge parameter q on Telegram (ablation) | DSHN · Telegram | accuracy | Improved |
| Increase number of layers (ablation) | DSHN · node classification | accuracy | Improved |
| Increase number of layers (ablation) | HGNN · node classification | accuracy | Hurt |
| Increase stalk dimension d (ablation) | DSHN · node classification | accuracy | Improved |
| Detach gradients through Laplacian construction (DSHNLight vs DSHN) (ablation) | DSHN · Squirrel | accuracy: 43.55 -> 44.09 | No clear effect |
| Use DSHNLight instead of DSHN on Telegram (ablation) | DSHN · Telegram | accuracy: 88.65 -> 81.15 | Hurt |
| Use DSHNLight instead of DSHN on Roman-empire (ablation) | DSHN · Roman-empire | accuracy: OOM -> 89.24 | Improved |
Closest papers by meaning
- Sheaf Hypergraph Networks (NeurIPS 2023 · similarity 0.95)
- Mitigating Over-squashing without Rewiring: A Sheaf Effective Resistance Perspective (NeurIPS 2026 · similarity 0.88)
- When Hypergraph Meets Heterophily: New Benchmark Datasets and Baseline (AAAI 2025 · similarity 0.88)
- SPHINX: Structural Prediction using Hypergraph Inference Network (ICML 2025 · similarity 0.88)
- Beyond Graphs: Can Large Language Models Comprehend Hypergraphs? (ICLR 2025 · similarity 0.87)
- HypeBoy: Generative Self-Supervised Representation Learning on Hypergraphs (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.
