Sampling from multi-modal distributions with polynomial query complexity in fixed dimension via reverse diffusion
Adrien Vacher, Omar Chehab, Anna Korba · 2025
Read Sampling from multi-modal distributions with polynomial query complexity in fixed dimension via reverse diffusion by Adrien Vacher, Omar Chehab, Anna Korba (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-2501-00565.txt.
Description
Adrien Vacher, Omar Chehab, and Anna Korba present a sampling method in a paper published in 2024. Drawing fair samples from a distribution with many peaks is a basic hard problem in statistics and machine learning. Peaks can trap older methods, and many methods need to know where the peaks are in advance.
The new method runs a diffusion process backward in time, estimating the guiding scores with a self-adjusting Monte Carlo step. It needs no advance knowledge of peak locations and avoids getting stuck between them. The cost of the method grows in a controlled, polynomial way with the traits that make a distribution many-peaked. The guarantee holds in fixed dimension and covers broad cases, including all mixtures of bell curves. That last case had been shut out by older assumptions, so the result widens what can be sampled with confidence.
Catalog record
| CREATOR | Adrien Vacher, Omar Chehab, Anna Korba |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 102 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-00565