Research map / Graph learning / Graph self-supervised learning
Beltrami Flow and Neural Diffusion on Graphs
Benjamin Paul Chamberlain, James Rowbottom, Davide Eynard, Francesco Di Giovanni, Xiaowen Dong, Michael M. Bronstein
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 scaled-dot attention instead of cosine similarity (ablation) | BLEND · Cora | test accuracy: 83.8 -> 84.8 | Improved |
| use scaled-dot attention instead of Pearson correlation attention (ablation) | BLEND · Pubmed | test accuracy: 80.0 -> 79.5 | Hurt |
| vary attention function (scaled dot, cosine, Pearson, exp kernel) (ablation) | BLEND · Citeseer | test accuracy: 73.5 to 75.9 | No clear effect |
| add positional encodings (ablation) | BLEND · Cora/Citeseer/Pubmed etc. | test accuracy | Improved |
| add positional encodings (Coauthor CS) (ablation) | BLEND · Coauthor CS | test accuracy | No clear effect |
| use DIGL PPR positional encodings instead of hyperbolic embeddings (ablation) | BLEND · six datasets average | test accuracy: 85.28 -> 85.48 | No clear effect |
| use BLEND instead of GAT | BLEND · Cora random splits | test accuracy: 81.8 -> 84.8 | Improved |
| use BLEND instead of GAT | BLEND · ogb-arxiv | test accuracy: 73.01 -> 72.56 | Hurt |
| use kNN rewiring in positional space (BLEND-kNN) | BLEND · Pubmed random splits | test accuracy: 79.5 -> 80.9 | Improved |
| use kNN rewiring in positional space (BLEND-kNN) | BLEND · Cora random splits | test accuracy: 84.8 -> 82.5 | Hurt |
| use adaptive Dopri5 solver instead of fixed-step Euler (ablation) | BLEND · six datasets | test accuracy | Improved |
Closest papers by meaning
- GRAND: Graph Neural Diffusion (ICML 2021 · similarity 0.89)
- Bundle Neural Network for message diffusion on graphs (ICLR 2025 · similarity 0.88)
- A Generalized Neural Diffusion Framework on Graphs (AAAI 2024 · similarity 0.87)
- Hyperbolic-PDE GNN: Spectral Graph Neural Networks in the Perspective of A System of Hyperbolic Partial Differential Equations (ICML 2025 · similarity 0.86)
- Feature Transportation Improves Graph Neural Networks (AAAI 2024 · similarity 0.86)
- Permutation Equivariant Neural Controlled Differential Equations for Dynamic Graph Representation Learning (NeurIPS 2025 · similarity 0.85)
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.
