Grid Particle Gibbs with Ancestor Sampling for State-Space Models

Mary Llewellyn, Ruth King, Víctor Elvira, Gordon Ross · 2025

Read Grid Particle Gibbs with Ancestor Sampling for State-Space Models by Mary Llewellyn, Ruth King, Víctor Elvira, Gordon Ross (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-2501-03395.txt.

Description

Mary Llewellyn and co-authors speed up a Bayesian workhorse for hidden-state models by laying a grid under its particles. State-space models infer unseen states and parameters together, and particle Gibbs samplers do it well but slowly. Their 2025 algorithm replaces random particle proposals with a deterministic grid read as an approximate hidden Markov model.

Ancestor sampling keeps the particle family tree healthy, and a set of practical tweaks trims the remaining waste. The test beds are deliberately nasty: regime-switching models, including one fit to tourism demand after COVID scrambled it. Against earlier particle Gibbs versions, the gains are substantial. Statisticians modeling epidemics, economies, or animal movement get Bayesian rigor at a friendlier clock speed.

Catalog record

CREATORMary Llewellyn, Ruth King, Víctor Elvira, Gordon Ross
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE69 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-03395

Browse the full library · About the Tower of Babel