Research map / Learning theory and algorithms / Optimal transport
Multiscale Strategies for Computing Optimal Transport
Samuel Gerber, Mauro Maggioni
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 multiscale transport coordinates instead of single-scale (ablation) | OT distance embedding · OASIS MMSE | R^2: 0.17 -> 0.30 | Improved |
| use multiscale instead of single-scale transport (ablation) | OT distance embedding · OASIS age | R^2: 0.78 -> 0.82 | Improved |
| use multiscale OT instead of LDDMM | OT distance embedding · OASIS MMSE | R^2: 0.18 -> 0.30 | Improved |
| use multiscale OT instead of LDDMM | OT distance embedding · OASIS CDR | R^2: 0.36 -> 0.38 | Improved |
| add refinement (neighborhood radius factor above 1) to capacity propagation (ablation) | multiscale OT · ellipse and Caffarelli, 5000 points | relative error | No clear effect |
| add iterated capacity propagation instead of simple propagation (ablation) | multiscale OT · ellipse and Caffarelli, 5000 points | relative error: errors under one percent | Improved |
| combine potential refinement with capacity propagation (ablation) | multiscale OT · ellipse and Caffarelli, 5000 points | relative error and time | Improved |
| use multiscale instead of network simplex or Sinkhorn | multiscale OT · ellipse, growing point count | computation time: memory limit around 2·10^4 constraints vs 2·10^6 constraints | Improved |
| reduce Sinkhorn tolerance from 1e-5 to 1e-2 (ablation) | Sinkhorn · ellipse, growing point count | computation time: approximately 10% speedup | No clear effect |
| use potential refinement when transport distance is large relative to support | multiscale OT · uniform squares | relative error | Hurt |
Closest papers by meaning
- Geometry-Aware Optimal Transport: Fast Intrinsic Dimension and Wasserstein Distance Estimation (NeurIPS 2026 · similarity 0.87)
- Computing Optimal Transport Maps and Wasserstein Barycenters Using Conditional Normalizing Flows (ICML 2025 · similarity 0.86)
- Linear Spherical Sliced Optimal Transport: A Fast Metric for Comparing Spherical Data (ICLR 2025 · similarity 0.86)
- Hierarchical Refinement: Optimal Transport to Infinity and Beyond (ICML 2025 · similarity 0.86)
- A Truncated Newton Method for Optimal Transport (ICLR 2025 · similarity 0.86)
- On Partial Optimal Transport: Revising the Infeasibility of Sinkhorn and Efficient Gradient Methods (AAAI 2024 · similarity 0.85)
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.
