Research map / Graph learning / Graph self-supervised learning
Pseudo-Riemannian Graph Convolutional 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 |
|---|---|---|---|
| replace Riemannian (hyperbolic/product) manifold with pseudo-hyperboloid Q^{7,3} (ablation) | GCN embedding · Web-Edu graph reconstruction | mAP: 89.21 -> 99.67 | Helped |
| use pseudo-hyperboloid Q^{7,3} instead of best Riemannian baseline (ablation) | GCN embedding · Power graph reconstruction | mAP: 95.82 -> 100.00 | Helped |
| use Q^{3,7} instead of best product-manifold baseline (ablation) | GCN embedding · Facebook graph reconstruction | mAP: 87.73 -> 91.75 | Helped |
| vary time dimension from t=1 (hyperboloid) to t=3 on Q-GCN (ablation) | Q-GCN · Web-Edu graph reconstruction | mAP: 87.03 -> 99.67 | Helped |
| use all-time dimensions (sphere-like, t=10) instead of intermediate t (ablation) | Q-GCN · Web-Edu graph reconstruction | mAP: 99.67 -> 82.57 | Hurt |
| use more time dimensions on low-hierarchy graph (Facebook) (ablation) | Q-GCN · Facebook graph reconstruction | mAP: 81.60 (t=1) -> 91.75 (t=7) -> 83.81 (t=10) | Mixed |
| pseudo-hyperboloid Q-GCN versus hyperbolic HGCN on tree-like graphs (ablation) | Q-GCN · Airport link prediction | ROC AUC: 96.03 vs 96.30 | No clear effect |
| pseudo-hyperboloid with many time dimensions versus best Riemannian baseline (ablation) | Q-GCN · Cora link prediction | ROC AUC: 94.04 -> 95.16 | Helped |
| pseudo-hyperboloid with Q^{1,15} versus best Riemannian baseline (ablation) | Q-GCN · Cora node classification | F1: 83.05 -> 83.72 | Helped |
| replace hyperbolic HGCN by Q^{14,2} (ablation) | Q-GCN · Pubmed node classification | F1: 80.50 -> 81.34 | Helped |
| add neighborhood aggregation (Q-GCN vs Q-NN) (ablation) | Q-GCN · node classification | F1 | Helped |
Closest papers by meaning
- Pseudo-Riemannian Graph Transformer (NeurIPS 2025 · similarity 0.93)
- Toward a Unified Geometry Understanding : Riemannian Diffusion Framework for Graph Generation and Prediction (NeurIPS 2025 · similarity 0.90)
- Building Transformation Layers for Riemannian Neural Networks (NeurIPS 2026 · similarity 0.89)
- Spectro-Riemannian Graph Neural Networks (ICLR 2025 · similarity 0.89)
- Deeper with Riemannian Geometry: Overcoming Oversmoothing and Oversquashing for Graph Foundation Models (NeurIPS 2025 · similarity 0.88)
- Motif-Aware Riemannian Graph Neural Network with Generative-Contrastive Learning (AAAI 2024 · 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.