Dynamic Diffusion Schrödinger Bridge in Astrophysical Observational Inversions
Ye Zhu, Duo Xu, Zhiwei Deng, Jonathan C. Tan, Olga Russakovsky · 2025
Read Dynamic Diffusion Schrödinger Bridge in Astrophysical Observational Inversions by Ye Zhu, Duo Xu, Zhiwei Deng, Jonathan C. Tan, Olga Russakovsky (2025) 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-2506-08065.txt.
Description
Ye Zhu, Duo Xu, Zhiwei Deng, Jonathan C. Tan, and Olga Russakovsky adapt a generative model to read the physics of star-forming clouds, in a 2025 paper. Giant molecular clouds are where stars are born. Observers must infer hidden conditions from limited views. The team built a variant of a diffusion bridge model. It learns to map between paired states of a system over time. They tested it on simulated clouds and on real data from the Taurus region.
From the astronomy side, the paired method beat standard statistics and other machine learning tools. It was clearer to interpret and faster to train. From the modeling side, the generative approach beat direct pixel prediction. The edge showed most on test cases with unseen starting states. Different physics driving those cases did not break it. The study argues generative models can learn real dynamics, not just data patterns.
Catalog record
| CREATOR | Ye Zhu, Duo Xu, Zhiwei Deng, Jonathan C. Tan, Olga Russakovsky |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 55 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-08065