Adaptive tuning of Hamiltonian Monte Carlo methods

Elena Akhmatskaya, Lorenzo Nagar, Jose Antonio Carrillo, Leonardo Gavira Balmacz, Hristo Inouzhe, Martín Parga Pazos, María Xosé Rodríguez Álvarez · 2025

Read Adaptive tuning of Hamiltonian Monte Carlo methods by Elena Akhmatskaya, Lorenzo Nagar, Jose Antonio Carrillo, Leonardo Gavira Balmacz, Hristo Inouzhe, Martín Parga Pazos, María Xosé Rodríguez Álvarez (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-2506-04082.txt.

Description

Elena Akhmatskaya and co-authors present a way to tune a popular sampling method automatically, in a 2025 paper. The method, Hamiltonian Monte Carlo, draws samples from complex probability models by mimicking physics motion. Its speed and accuracy hinge on settings that users usually pick by rule of thumb.

Their approach, called ATune, studies a simple bell-curve model in theory and mixes in data from a short warm-up run. From that, it picks a sound numerical solver and safe ranges for the sampler's settings, cutting values that cause wild errors or poor sampling. Tests on standard models and real cases, from cancer therapy to flu spread, beat hand-tuned versions and a leading rival sampler. The software adds no cost to the main run.

Catalog record

CREATORElena Akhmatskaya, Lorenzo Nagar, Jose Antonio Carrillo, Leonardo Gavira Balmacz, Hristo Inouzhe, Martín Parga Pazos, María Xosé Rodríguez Álvarez
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE102 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2506-04082

Browse the full library · About the Tower of Babel