Osmundsen, Kjartan Kloster, Kleppe, Tore Selland and Liesenfeld, Roman ORCID: 0000-0001-6996-6215 (2021). Importance Sampling-Based Transport Map Hamiltonian Monte Carlo for Bayesian Hierarchical Models. J. Comput. Graph. Stat., 30 (4). S. 906 - 920. PHILADELPHIA: TAYLOR & FRANCIS INC. ISSN 1537-2715

Full text not available from this repository.

Abstract

We propose an importance sampling (IS)-based transport map Hamiltonian Monte Carlo procedure for performing a Bayesian analysis in nonlinear high-dimensional hierarchical models. Using IS techniques to construct a transport map, the proposed method transforms the typically highly complex posterior distribution of a hierarchical model such that it can be easily sampled using standard HamiltonianMonte Carlo. In contrast to standard applications of high-dimensional IS, our approach does not require IS distributions with high fidelity, whichmakes it computationally very cheap. Moreover, it is less prone to the notorious problem of IS that the variance of IS weights can become infinite. We illustrate our algorithm with applications to challenging dynamic state-space models, where it exhibits very high simulation efficiency compared to relevant benchmarks, even for variants of the proposed method implemented using a few dozen lines of code in the Stan statistical software. The article is accompanied by supplementary material containing further details, and the computer code is available at https://github.com/kjartako/TMHMC. These are also supplementary materials for this article are available online.

Item Type: Journal Article
Creators:
CreatorsEmailORCIDORCID Put Code
Osmundsen, Kjartan KlosterUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Kleppe, Tore SellandUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Liesenfeld, RomanUNSPECIFIEDorcid.org/0000-0001-6996-6215UNSPECIFIED
URN: urn:nbn:de:hbz:38-570196
DOI: 10.1080/10618600.2021.1923519
Journal or Publication Title: J. Comput. Graph. Stat.
Volume: 30
Number: 4
Page Range: S. 906 - 920
Date: 2021
Publisher: TAYLOR & FRANCIS INC
Place of Publication: PHILADELPHIA
ISSN: 1537-2715
Language: English
Faculty: Unspecified
Divisions: Unspecified
Subjects: no entry
Uncontrolled Keywords:
KeywordsLanguage
STOCHASTIC VOLATILITY; INFERENCEMultiple languages
Statistics & ProbabilityMultiple languages
URI: http://kups.ub.uni-koeln.de/id/eprint/57019

Downloads

Downloads per month over past year

Altmetric

Export

Actions (login required)

View Item View Item