Rate-Optimal Noise Annealing in Semi-Dual Neural Optimal Transport: Tangential Identifiability, Off-Manifold Ambiguity, and Guaranteed Recovery
Raymond Chu, Jaewoong Choi, Dohyun Kwon · 2026
Read Rate-Optimal Noise Annealing in Semi-Dual Neural Optimal Transport: Tangential Identifiability, Off-Manifold Ambiguity, and Guaranteed Recovery by Raymond Chu, Jaewoong Choi, Dohyun Kwon (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-04110.txt.
Description
Optimal transport models can learn the wrong map and never notice. Raymond Chu and 2 co-authors explain when that happens in semi-dual neural optimal transport. Their 2026 paper studies data concentrated on low-dimensional manifolds, where the training objective constrains the map on the data but leaves it ambiguous off the manifold.
Adding noise during training fixes the ambiguity, if annealed correctly. The authors derive a computable final noise level achieving the optimal statistical rate, governed by the data's intrinsic dimension. Their analysis unifies stability of transport plans, smoothing bias, and finite-sample error. Practitioners get a formula, not a guess, for a critical hyperparameter.
Catalog record
| CREATOR | Raymond Chu, Jaewoong Choi, Dohyun Kwon |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 85 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-04110