Investigating the applicability of parallelized straight simulation

Keywords

Loading...
Thumbnail Image

Issue Date

2022-06-20

Language

en

Document type

Journal Title

Journal ISSN

Volume Title

Publisher

Title

ISSN

Volume

Issue

Startpage

Endpage

DOI

Abstract

Straight simulation is an approximate Bayesian inference algorithm and an instance of the Markov Chain Monte Carlo algorithms. It has been developed to resolve some issues of prior techniques but ended up showing some problems in other areas itself. As part of this project the applicability of parallelizing straight simulation will be investigated, since this might enable a significant performance increase. In particular the influencing properties for poor or decent accuracy will be explored through exploratory testing. Nonetheless, completely parallelizing straight simulation generally implies the loss of guarantee to converge to the posterior distribution. The discussion of the results support this observation in most cases, while describing under which circumstance a decent approximation is still viable for parallel straight simulation.

Description

Citation

Faculty

Faculteit der Sociale Wetenschappen