Research map / Learning theory and algorithms / Optimal transport
Fast Optimal Transport through Sliced Generalized Wasserstein Geodesics
Guillaume Mahey, Laetitia Chapel, Gilles Gasso, Clément Bonet, Nicolas Courty
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 |
|---|---|---|---|
| Use min-SWGG transport map instead of nearest-neighbor correspondence in ICP | ICP · Point clouds n=500 | Sinkhorn divergence: 3.54 -> 0.05 | Improved |
| Use min-SWGG transport map instead of nearest-neighbor correspondence in ICP | ICP · Point clouds n=3000 | Sinkhorn divergence: 96.9 -> 37.6 | Improved |
| Use min-SWGG transport map instead of nearest-neighbor correspondence in ICP | ICP · Point clouds n=150000 | Sinkhorn divergence: 23.3 -> 6.7 | Improved |
| Use min-SWGG instead of exact OT map in ICP | ICP · Point clouds n=500 | Sinkhorn divergence: 0.32 -> 0.05 | Improved |
| Use min-SWGG instead of exact OT map in ICP | ICP · Point clouds n=3000 | Sinkhorn divergence: 48.4 -> 37.6 | Improved |
| Use min-SWGG instead of exact OT map in ICP | ICP · Point clouds n=3000 | running time (s): 58.46 -> 0.90 | Improved |
| Use min-SWGG instead of exact OT map in ICP | ICP · Point clouds n=500 | running time (s): 0.18 -> 0.04 is faster; NN is 0.02 | Hurt |
| Use min-SWGG instead of NN correspondence in ICP | ICP · Point clouds n=150000 | running time (s): 59.37 -> 105.75 | Hurt |
| Use min-SWGG when exact OT is infeasible | ICP · Point clouds n=150000 | feasibility | Improved |
| Optimize projection direction by gradient descent instead of random search (ablation) | min-SWGG · Gaussians d=200, n=1000 | approximation of WD | Improved |
| Optimize projection direction vs random search in low dimension (ablation) | min-SWGG · Gaussians d=2 | approximation of WD | No clear effect |
| Use gradient-descent min-SWGG vs SW, max-SW, PWD in gradient flow | min-SWGG · Gradient flow d=500, n=100 | alignment of distributions | Improved |
Closest papers by meaning
- Differentiable Generalized Sliced Wasserstein Plans (NeurIPS 2025 · similarity 0.92)
- Tree-Sliced Wasserstein Distance: A Geometric Perspective (ICML 2025 · similarity 0.91)
- A Provably Convergent and Practical Algorithm for Gromov–Wasserstein Optimal Transport (NeurIPS 2026 · similarity 0.91)
- Distance-Based Tree-Sliced Wasserstein Distance (ICLR 2025 · similarity 0.89)
- Expected Sliced Transport Plans (ICLR 2025 · similarity 0.89)
- Towards Understanding Gradient Dynamics of the Sliced-Wasserstein Distance via Critical Point Analysis (ICML 2025 · similarity 0.89)
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.
