Score-based Metropolis-Hastings for Fractional Langevin Algorithms
Ahmed Aloui [email protected], Junyi Liao [email protected], Ali Hasan [email protected], Jose Blanchet [email protected], Vahid Tarokh [email protected] · 2026
Read Score-based Metropolis-Hastings for Fractional Langevin Algorithms by Ahmed Aloui [email protected], Junyi Liao [email protected], Ali Hasan [email protected], Jose Blanchet [email protected], Vahid Tarokh [email protected] (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-2602-00835.txt.
Description
Ahmed Aloui and colleagues introduce the Metropolis-Adjusted Fractional Langevin Algorithm, a score-based fix for sampling heavy-tailed, multimodal targets. Older methods fail here because neither target nor proposal density can be evaluated. Older methods run unadjusted and can show large finite-time errors with poor tail control. MAFLA uses proxies for proposal scores under alpha-stable noise and derives an acceptance function through Score Balance Matching. Tests on combinatorial tasks show it strongly improves finite-time sampling over older dynamics.
Catalog record
| CREATOR | Ahmed Aloui [email protected], Junyi Liao [email protected], Ali Hasan [email protected], Jose Blanchet [email protected], Vahid Tarokh [email protected] |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 59161 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602.00835