Research map / Learning theory and algorithms / Optimal transport
Wasserstein Distances Made Explainable: Insights Into Dataset Shifts and Transport Phenomena
Philip Naumann, Jacob Kauffmann, Grégoire Montavon
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 WaX instead of Occlusion at high q | Wasserstein explanation · Robot (W2, q=inf) | SRG: 0.72 -> 1.14 | Improved |
| use WaX instead of MeanShift at high p | Wasserstein explanation · Robot (W10, q=2) | SRG: 4.81 -> 8.33 | Improved |
| add alpha and beta tuning to Coupling (WaX instead of Coupling) (ablation) | Coupling baseline · Robot (W10, q=2) | SRG: -0.68 -> 8.33 | Improved |
| use WaX instead of Coupling at q=inf | Coupling baseline · Crime (W2, q=inf) | SRG: 0.25 -> 0.59 | Improved |
| use WaX instead of Occlusion at low q (equal accuracy, cheaper) | Occlusion · Robot (W1, q=1) | SRG: 9.10 -> 9.11 | No clear effect |
| use WaX instead of MeanShift on Robot W1 | MeanShift · Robot (W1, q=2) | SRG: 0.06 -> 1.48 | Improved |
| use exact coupling instead of uniform coupling in WaX (ablation) | WaX · Air Quality (every 1h) | cosine similarity: 0.80 -> 0.85 | Improved |
| use uniform coupling instead of exact coupling in WaX (ablation) | WaX · Electricity (every 1h) | cosine similarity: 0.83 -> 0.88 | Improved |
| use WaX instead of Logistic [GI] baseline | Logistic · Air Quality (every 1h) | cosine similarity: 0.47 -> 0.85 | Improved |
| use WaX instead of Uniform relevance | Uniform · Appliances (every 1h) | cosine similarity: 0.38 -> 0.51 | Improved |
| use WaX instead of Uniform on Electricity | Uniform · Electricity (every 6h) | cosine similarity: 0.94 -> 0.98 | No clear effect |
| use WaX exact coupling instead of MeanShift on PLISM UNI | MeanShift · PLISM (UNI) | cosine similarity: 0.77 -> 0.99 | Improved |
Closest papers by meaning
- Tree-Wasserstein Distance for High Dimensional Data with a Latent Feature Hierarchy (ICLR 2025 · similarity 0.87)
- A New Robust Partial p-Wasserstein-Based Metric for Comparing Distributions (ICML 2024 · similarity 0.86)
- Flowing Datasets with Wasserstein over Wasserstein Gradient Flows (ICML 2025 · similarity 0.86)
- Statistical and Computational Guarantees of Kernel Max-Sliced Wasserstein Distances (ICML 2025 · similarity 0.86)
- Sliced Wasserstein Estimation with Control Variates (ICLR 2024 · similarity 0.86)
- Wasserstein Wormhole: Scalable Optimal Transport Distance with Transformer (ICML 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.
