Research map / Graph learning / GNN architectures
Neural Sheaf Diffusion: A Topological Perspective on Heterophily and Oversmoothing in GNNs
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 |
|---|---|---|---|
| learn general (asymmetric) sheaf maps instead of symmetric maps (weighted graph Laplacian) (ablation) | sheaf diffusion, d=1 · synthetic bipartite graph | linear separability of classes | Helped |
| use sheaf diffusion with learned restriction maps instead of GCN-style graph convolution on heterophilic graphs | NSD · Texas | accuracy (%): 55.14 -> 85.95 | Helped |
| use sheaf diffusion instead of GCN | NSD · Cornell | accuracy (%): 60.54 -> 86.49 | Helped |
| NSD versus GCN on homophilic Cora | NSD · Cora | accuracy (%): 86.98 -> 87.14 | No clear effect |
| orthogonal restriction maps versus diagonal or general maps (ablation) | NSD · 9 node classification datasets | accuracy (%): Texas 85.95 vs 82.97 (Gen) | Mixed |
| general (unconstrained) restriction maps instead of orthogonal (ablation) | NSD · Texas | accuracy (%): 85.95 -> 82.97 | Hurt |
Closest papers by meaning
- Mitigating Over-squashing without Rewiring: A Sheaf Effective Resistance Perspective (NeurIPS 2026 · similarity 0.91)
- A Generalized Neural Diffusion Framework on Graphs (AAAI 2024 · similarity 0.91)
- Bundle Neural Network for message diffusion on graphs (ICLR 2025 · similarity 0.89)
- Hyperbolic-PDE GNN: Spectral Graph Neural Networks in the Perspective of A System of Hyperbolic Partial Differential Equations (ICML 2025 · similarity 0.87)
- Understanding Heterophily for Graph Neural Networks (ICML 2024 · similarity 0.87)
- Hyperbolic Geometric Latent Diffusion Model for Graph Generation (ICML 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.