Weighted Stochastic Differential Equation to Implement Wasserstein-Fisher-Rao Gradient Flow
2025
Read Weighted Stochastic Differential Equation to Implement Wasserstein-Fisher-Rao Gradient Flow (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-2512-17878.txt.
Introduction
Herlock Rahimi studies a math problem behind AI image makers: how to draw samples from a complex target shape. Standard diffusion methods wander slowly when the target has several separate humps. A richer geometry, named Wasserstein-Fisher-Rao, lets mass grow or shrink as well as move, which could speed the search.
Rahimi shows how to add the needed correction terms through weighted random equations, using a classic Feynman-Kac form. The study is early and careful rather than a finished tool. It maps the shape and operator structure of these flows, giving later work a firm base for faster samplers.
Catalog record
| YEAR | 2025 |
|---|---|
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 133 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2512-17878