Research map / Probabilistic and scientific modeling / Variational and generative sampling
CO-SNE: Dimensionality Reduction and Visualization for Hyperbolic Data
Check 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 CO-SNE instead of HoroPCA to reduce hyperbolic neural network features to two dimensions | Hyperbolic neural network features · MNIST sampled test images (100 per class) | accuracy: 30.2% -> 61.2% | Helped |
| Remove the distance loss (set λ₂ = 0) (ablation) | CO-SNE · Synthetic five-cluster hyperbolic data | preservation of hierarchical structure | Hurt |
| Remove the KL-divergence loss (set λ₁ = 0) (ablation) | CO-SNE · Synthetic five-cluster hyperbolic data | preservation of local similarity | Hurt |
| Increase the distance-loss weight λ₂ (ablation) | CO-SNE · Synthetic five-cluster hyperbolic data | preservation of hierarchical structure | Helped |
| Increase the distance-loss weight λ₂ (ablation) | CO-SNE · Synthetic five-cluster hyperbolic data | preservation of local similarity | Hurt |
| Use a larger KL-divergence weight λ₁ (ablation) | CO-SNE · Synthetic five-cluster hyperbolic data | visualization quality | Hurt |
| Use hyperbolic Cauchy rather than hyperbolic Student's t for low-dimensional similarities (ablation) | Hyperbolic t-SNE (HT-SNE) · Synthetic hyperbolic point clusters | repulsion between dissimilar points | Helped |
| Add the distance loss to preserve each point's distance from the origin (ablation) | CO-SNE · Synthetic five-cluster hyperbolic data | preservation of hierarchical structure | Helped |
Closest papers by meaning
- Hyperbolic Representation Learning: Revisiting and Advancing (ICML 2023 · similarity 0.90)
- Learning Continuous Hierarchies in the Lorentz Model of Hyperbolic Geometry (ICML 2018 · similarity 0.89)
- Representation Tradeoffs for Hyperbolic Embeddings (ICML 2018 · similarity 0.89)
- Hyperbolic Embeddings of Supervised Models (NeurIPS 2024 · similarity 0.88)
- Finsler Multi-Dimensional Scaling: Manifold Learning for Asymmetric Dimensionality Reduction and Embedding (CVPR 2025 · similarity 0.86)
- Unsupervised Feature Learning with Emergent Data-Driven Prototypicality (CVPR 2024 · 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.