Research map / Learning theory and algorithms / Optimal transport
Geometric Operator Learning with Optimal Transport
Xinyi Li, Zongyi Li, Nikola Kovachki, Anima Anandkumar
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 Mean strategy (nearest to transported mesh) rather than Max (ablation) | OTNO(Plan) · ShapeNet Car | relative L2: 7.01% -> 6.70% | Improved |
| use Mean strategy rather than Max (ablation) | OTNO(Plan) · DrivAerNet | MSE (e-5): 4.8 -> 3.4 | Improved |
| use single closest index rather than multi-encoder top-k (ablation) | OTNO(Plan) · ShapeNet Car | relative L2: 20.55% -> 6.70% | Improved |
| use single closest index rather than multi-decoder top-k (ablation) | OTNO(Plan) · ShapeNet Car | relative L2: 72.30% -> 6.70% | Improved |
| add cross-product normal features (ablation) | OTNO(Plan) · ShapeNet Car | relative L2: 7.19% -> 6.70% | Improved |
| add cross-product normal features (ablation) | OTNO(Plan) · DrivAerNet | MSE (e-5): 3.4 -> 3.4 | No clear effect |
| use torus latent mesh instead of sphere (ablation) | OTNO(Plan) · ShapeNet Car | relative L2: 7.09% -> 6.70% | Improved |
| use torus latent mesh instead of sphere (ablation) | OTNO(Plan) · DrivAerNet | MSE (e-5): 4.1 -> 3.4 | Improved |
| use double sphere instead of sphere latent mesh (ablation) | OTNO(Plan) · ShapeNet Car | relative L2: 7.09% -> 7.41% | Hurt |
| expand latent mesh by factor 3 (ablation) | OTNO(Plan) · ShapeNet and DrivAerNet | accuracy | Improved |
| replace conformal mapping with optimal transport (ablation) | OTNO · ShapeNet Car | relative L2: 12.79% -> 6.70% | Improved |
| use Plan (Sinkhorn) instead of Map (PPMM) (ablation) | OTNO · DrivAerNet | MSE and time | Improved |
Closest papers by meaning
- Geometry Aware Operator Transformer as an efficient and accurate neural surrogate for PDEs on arbitrary domains (NeurIPS 2025 · similarity 0.89)
- GAIA: Geometry-Adaptive Operator Learning for Forward and Inverse Problems (NeurIPS 2026 · similarity 0.87)
- Data-Efficient Operator Learning via Unsupervised Pretraining and In-Context Learning (NeurIPS 2024 · similarity 0.87)
- Reference Neural Operators: Learning the Smooth Dependence of Solutions of PDEs on Geometric Deformations (ICML 2024 · similarity 0.87)
- UGM2N: An Unsupervised and Generalizable Mesh Movement Network via M-Uniform Loss (NeurIPS 2025 · similarity 0.87)
- RIGNO: A Graph-based Framework For Robust And Accurate Operator Learning For PDEs On Arbitrary Domains (NeurIPS 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.
