Research map / Graph learning / GNN architectures
Hyperbolic Graph Neural Networks
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 |
|---|---|---|---|
| Use hyperbolic geometry instead of Euclidean geometry | HGNN graph-generation classifier · Synthetic graph generation classification, low-dimensional settings | macro F1: 77.2 $\pm$ 0.12 -> 93.0 $\pm$ 0.05; 77.2 $\pm$ 0.12 -> 94.1 $\pm$ 0.03 | Helped |
| Use hyperbolic geometry instead of Euclidean geometry | HGNN graph-generation classifier · Synthetic graph generation classification, 256 dimensions | macro F1: 95.3 $\pm$ 0.17 -> 93.7 $\pm$ 0.05; 95.3 $\pm$ 0.17 -> 95.3 $\pm$ 0.28 | Mixed |
| Use hyperbolic rather than Euclidean geometry | HGNN molecular property predictor · ZINC logP, QED, and SAS | MAE (scaled by 100): 3.3 $\pm$ 0.00 -> 2.4 $\pm$ 0.02; 6.4 $\pm$ 0.06 -> 4.7 $\pm$ 0.00; 9.2 $\pm$ 0.08 -> 7.7 $\pm$ 0.06 | Helped |
| Use Lorentz rather than Poincare geometry | HGNN molecular property predictor · ZINC logP, QED, and SAS | MAE (scaled by 100): 3.1 $\pm$ 0.01 -> 2.4 $\pm$ 0.02; 6.0 $\pm$ 0.04 -> 4.7 $\pm$ 0.00; 8.6 $\pm$ 0.02 -> 7.7 $\pm$ 0.06 | Helped |
| Use Lorentz HGNN rather than the strongest listed non-HGNN baseline, GGNN | ZINC molecular property prediction · ZINC logP, QED, and SAS | MAE (scaled by 100): 3.2 $\pm$ 0.20 -> 2.4 $\pm$ 0.02; 6.4 $\pm$ 0.20 -> 4.7 $\pm$ 0.00; 9.1 $\pm$ 0.10 -> 7.7 $\pm$ 0.06 | Helped |
| Use hyperbolic geometry rather than Euclidean geometry | HGNN price-fluctuation predictor · Ethereum Ether/USDT, Dev and Test | accuracy: 56.15 $\pm$ 0.30 -> 57.52 $\pm$ 0.35; 53.95 $\pm$ 0.20 -> 55.51 $\pm$ 0.37 | Helped |
| Use Lorentz rather than Poincare geometry | HGNN price-fluctuation predictor · Ethereum Ether/USDT, Dev and Test | accuracy: 57.03 $\pm$ 0.28 -> 57.52 $\pm$ 0.35; 54.41 $\pm$ 0.24 -> 55.51 $\pm$ 0.37 | Helped |
| Use hyperbolic geometry to represent graph structure | Graph-level classifier · Synthetic graph generation classification | macro F1 | Helped |
| Use hyperbolic representations for hierarchical transaction graphs | HGNN price-fluctuation predictor · Ethereum transaction graph | average node norm | Mixed |
Closest papers by meaning
- Hyperbolic Graph Convolutional Neural Networks (NeurIPS 2019 · similarity 0.91)
- Hyperbolic-PDE GNN: Spectral Graph Neural Networks in the Perspective of A System of Hyperbolic Partial Differential Equations (ICML 2025 · similarity 0.90)
- What do Graph Neural Networks learn? Insights from Tropical Geometry (NeurIPS 2024 · similarity 0.89)
- Geometry Awakening: Cross-Geometry Learning Exhibits Superiority over Individual Structures (NeurIPS 2024 · similarity 0.88)
- Weighted Embeddings for Low-Dimensional Graph Representation (AAAI 2025 · similarity 0.88)
- Residual Hyperbolic Graph Convolution Networks (AAAI 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.