Research map / Graph learning / Graph self-supervised learning
Hyperbolic Busemann Neural Networks
Ziheng Chen, Bernhard Schölkopf, Nicu Sebe
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 |
|---|---|---|---|
| replace Poincare MLR with BMLR-P | ResNet-18 head · ImageNet-1k val | top-1 accuracy (%): 71.77 -> 73.36 | Improved |
| replace Lorentz MLR with BMLR-L | ResNet-18 head · ImageNet-1k val | top-1 accuracy (%): 72.46 -> 73.24 | Improved |
| replace Poincare MLR with BMLR-P | ResNet-18 head · CIFAR-10 | top-1 accuracy (%): 95.04 -> 95.32 | No clear effect |
| replace Lorentz MLR with BMLR-L | ResNet-18 head · CIFAR-100 | top-1 accuracy (%): 78.03 -> 78.07 | No clear effect |
| use BMLR instead of Pseudo-Busemann MLR (compact parameters, batch efficient) | Poincare ResNet-18 head · ImageNet-1k val | fit time (s/epoch): 3907.12 -> 2300.77 | Improved |
| use BMLR-L instead of Lorentz MLR in genome CNN | CNN head · Genome Fungi (25 classes) | MCC: 15.07 -> 30.14 | Improved |
| use BMLR-P instead of PMLR on Covid variant classification | CNN head · Genome Covid (9 classes) | MCC: 74.09 -> 73.40 | Hurt |
| use BMLR-L instead of Lorentz MLR in HGCN | HGCN · Cora node classification | F1: 69.91 -> 77.65 | Improved |
| use prior hyperbolic MLR heads instead of tangent MLR on less hyperbolic graphs | HGCN · Cora, PubMed, Airport | F1: 77.37 -> 69.91 (LMLR on Cora) | Hurt |
| use BMLR-P instead of Poincare MLR in HGCN | HGCN · Disease node classification | F1: 88.98 -> 92.45 | Improved |
| replace Lorentz FC with BFC-L | hyperbolic NN link prediction · Disease link prediction | AUC: 72.78 -> 78.36 | Improved |
| replace Mobius FC with BFC-P | hyperbolic NN link prediction · PubMed link prediction | AUC: 94.93 -> 94.85 | No clear effect |
Closest papers by meaning
- Siegel Neural Networks (NeurIPS 2025 · similarity 0.87)
- Fully Hyperbolic Convolutional Neural Networks for Computer Vision (ICLR 2024 · similarity 0.86)
- Fast Hyperboloid Decision Tree Algorithms (ICLR 2024 · similarity 0.85)
- Riemannian Multinomial Logistics Regression for SPD Neural Networks (CVPR 2024 · similarity 0.85)
- RMLR: Extending Multinomial Logistic Regression into General Geometries (NeurIPS 2024 · similarity 0.85)
- Hyperbolic Dataset Distillation (NeurIPS 2025 · similarity 0.84)
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.
