Research map / Learning theory and algorithms / Prediction and decision losses
The Numerical Stability of Hyperbolic Representation Learning
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 Euclidean parametrization instead of the Poincaré or Lorentz model for tree embedding | Tree embedding · Simulated Trees 1–8 | average distortion: Tree 1: 1.0462 -> 1.0157; Tree 2: 1.0589 -> 1.0134 | Helped |
| Use Euclidean parametrization instead of the Lorentz model for tree embedding | Tree embedding · Simulated Trees 1–8 | average distortion: Tree 1: 1.0176 -> 1.0157; Tree 2: 1.0180 -> 1.0134 | Helped |
| Use Euclidean parametrization instead of the Poincaré model for tree embedding | Tree embedding · Simulated Trees 1–8 | average distortion: Tree 1: 1.0462 -> 1.0157; Tree 2: 1.0589 -> 1.0134 | Helped |
| Use LSVMPP instead of the strongest listed baseline | Hyperbolic SVM · CIFAR-10 | accuracy (%): 91.88 ± 0.00 -> 91.96 ± 0.00 | Helped |
| Use LSVMPP instead of the strongest listed baseline | Hyperbolic SVM · fashion-MNIST | accuracy (%): 86.57 ± 0.00 -> 89.49 ± 0.00 | Helped |
| Use LSVMPP instead of the strongest listed baseline | Hyperbolic SVM · Paul | accuracy (%): 58.36 ± 0.07 -> 62.64 ± 0.05 | Helped |
| Use arcsinh-scaled signed hyperplane distance for Platt scaling instead of raw Lorentz SVM output (ablation) | Lorentz SVM classification · Hyperbolic SVM datasets | accuracy | Helped |
Closest papers by meaning
- Lorentzian Distance Learning for Hyperbolic Representations (ICML 2019 · similarity 0.90)
- Robust large-margin learning in hyperbolic space (NeurIPS 2020 · similarity 0.90)
- Hyperbolic Neural Networks++ (ICLR 2021 · similarity 0.90)
- Hyperbolic Embeddings of Supervised Models (NeurIPS 2024 · similarity 0.88)
- Fast Hyperboloid Decision Tree Algorithms (ICLR 2024 · similarity 0.87)
- Hyperbolic Dataset Distillation (NeurIPS 2025 · similarity 0.86)
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.