Research map / Learning theory and algorithms / Optimal transport
Streaming Sliced Optimal Transport
Khai Nguyen
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 random sampling with quantile-sketch Stream-SW | SW estimation · Gaussians approximation error (n=10000, k=2) | relative error: 0.289892 -> 0.0989126 | Improved |
| replace random sampling with quantile-sketch Stream-SW at large sketch size | SW estimation · Gaussians approximation error (n=10000, k=1000) | relative error: 0.0420441 -> 0.0404598 | No clear effect |
| replace random sampling with Stream-SW | SW kNN classifier · ModelNet10 KNN (L=5, k=5) | accuracy (%): 39.33 -> 45.53 | Improved |
| replace random sampling with Stream-SW | SW kNN classifier · ModelNet10 KNN (L=100, k=10) | accuracy (%): 62 -> 76.67 | Improved |
| replace random sampling with Stream-SW at large sketch | SW kNN classifier · ModelNet10 KNN (L=5, k=50) | accuracy (%): 61.33 -> 68.67 | Improved |
| replace random sampling with Stream-SW | SW gradient flow · ShapeNet gradient flow (L=1000, k=200, step 5000) | Wasserstein-2 x10^3: 6.59 -> 0.48 | Improved |
| replace random sampling with Stream-SW | SW gradient flow · ShapeNet gradient flow (L=100, k=100, step 5000) | Wasserstein-2 x10^3: 9.36 -> 1.93 | Improved |
| use Stream-SW instead of Full SW (all samples) | SW gradient flow · ShapeNet gradient flow (L=100, step 5000) | Wasserstein-2 x10^3: 0.84 -> 1.36 | Hurt |
| replace sliding-window SW with Stream-SW | change point detection · MSRC-12 Guest 2 | detection delay: 100 -> 10 | Improved |
| replace sliding-window SW with Stream-SW | change point detection · MSRC-12 Guest 3 | detection delay: 49 -> 24 | Improved |
| increase sketch size k | Stream-SW · ShapeNet gradient flow (L=1000, step 5000) | Wasserstein-2 x10^3: 0.85 -> 0.48 | Improved |
Closest papers by meaning
- Differentiable Generalized Sliced Wasserstein Plans (NeurIPS 2025 · similarity 0.88)
- Distance-Based Tree-Sliced Wasserstein Distance (ICLR 2025 · similarity 0.88)
- Tree-Sliced Wasserstein Distance with Nonlinear Projection (ICML 2025 · similarity 0.87)
- Sliced Wasserstein Estimation with Control Variates (ICLR 2024 · similarity 0.87)
- Tree-Sliced Wasserstein Distance: A Geometric Perspective (ICML 2025 · similarity 0.87)
- Linear Spherical Sliced Optimal Transport: A Fast Metric for Comparing Spherical Data (ICLR 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.
