Tampere University of Technology

TUTCRIS Research Portal

A fast universal self-tuned sampler within Gibbs sampling

Research output: Contribution to journalArticleScientificpeer-review

Standard

A fast universal self-tuned sampler within Gibbs sampling. / Martino, L.; Yang, H.; Luengo, D.; Kanniainen, J.; Corander, J.

In: Digital Signal Processing, Vol. 47, 01.12.2015, p. 68-83.

Research output: Contribution to journalArticleScientificpeer-review

Harvard

Martino, L, Yang, H, Luengo, D, Kanniainen, J & Corander, J 2015, 'A fast universal self-tuned sampler within Gibbs sampling', Digital Signal Processing, vol. 47, pp. 68-83. https://doi.org/10.1016/j.dsp.2015.04.005

APA

Martino, L., Yang, H., Luengo, D., Kanniainen, J., & Corander, J. (2015). A fast universal self-tuned sampler within Gibbs sampling. Digital Signal Processing, 47, 68-83. https://doi.org/10.1016/j.dsp.2015.04.005

Vancouver

Martino L, Yang H, Luengo D, Kanniainen J, Corander J. A fast universal self-tuned sampler within Gibbs sampling. Digital Signal Processing. 2015 Dec 1;47:68-83. https://doi.org/10.1016/j.dsp.2015.04.005

Author

Martino, L. ; Yang, H. ; Luengo, D. ; Kanniainen, J. ; Corander, J. / A fast universal self-tuned sampler within Gibbs sampling. In: Digital Signal Processing. 2015 ; Vol. 47. pp. 68-83.

Bibtex - Download

@article{6bee9cfaba4c4548be48f50e78b65a31,
title = "A fast universal self-tuned sampler within Gibbs sampling",
abstract = "Bayesian inference often requires efficient numerical approximation algorithms, such as sequential Monte Carlo (SMC) and Markov chain Monte Carlo (MCMC) methods. The Gibbs sampler is a well-known MCMC technique, widely applied in many signal processing problems. Drawing samples from univariate full-conditional distributions efficiently is essential for the practical application of the Gibbs sampler. In this work, we present a simple, self-tuned and extremely efficient MCMC algorithm which produces virtually independent samples from these univariate target densities. The proposal density used is self-tuned and tailored to the specific target, but it is not adaptive. Instead, the proposal is adjusted during an initial optimization stage, following a simple and extremely effective procedure. Hence, we have named the newly proposed approach as FUSS (Fast Universal Self-tuned Sampler), as it can be used to sample from any bounded univariate distribution and also from any bounded multi-variate distribution, either directly or by embedding it within a Gibbs sampler. Numerical experiments, on several synthetic data sets (including a challenging parameter estimation problem in a chaotic system) and a high-dimensional financial signal processing problem, show its good performance in terms of speed and estimation accuracy.",
keywords = "Adaptive rejection Metropolis sampling, Bayesian inference, Gibbs sampling, Markov Chain Monte Carlo (MCMC), Metropolis within Gibbs",
author = "L. Martino and H. Yang and D. Luengo and J. Kanniainen and J. Corander",
year = "2015",
month = "12",
day = "1",
doi = "10.1016/j.dsp.2015.04.005",
language = "English",
volume = "47",
pages = "68--83",
journal = "Digital Signal Processing",
issn = "1051-2004",
publisher = "Elsevier",

}

RIS (suitable for import to EndNote) - Download

TY - JOUR

T1 - A fast universal self-tuned sampler within Gibbs sampling

AU - Martino, L.

AU - Yang, H.

AU - Luengo, D.

AU - Kanniainen, J.

AU - Corander, J.

PY - 2015/12/1

Y1 - 2015/12/1

N2 - Bayesian inference often requires efficient numerical approximation algorithms, such as sequential Monte Carlo (SMC) and Markov chain Monte Carlo (MCMC) methods. The Gibbs sampler is a well-known MCMC technique, widely applied in many signal processing problems. Drawing samples from univariate full-conditional distributions efficiently is essential for the practical application of the Gibbs sampler. In this work, we present a simple, self-tuned and extremely efficient MCMC algorithm which produces virtually independent samples from these univariate target densities. The proposal density used is self-tuned and tailored to the specific target, but it is not adaptive. Instead, the proposal is adjusted during an initial optimization stage, following a simple and extremely effective procedure. Hence, we have named the newly proposed approach as FUSS (Fast Universal Self-tuned Sampler), as it can be used to sample from any bounded univariate distribution and also from any bounded multi-variate distribution, either directly or by embedding it within a Gibbs sampler. Numerical experiments, on several synthetic data sets (including a challenging parameter estimation problem in a chaotic system) and a high-dimensional financial signal processing problem, show its good performance in terms of speed and estimation accuracy.

AB - Bayesian inference often requires efficient numerical approximation algorithms, such as sequential Monte Carlo (SMC) and Markov chain Monte Carlo (MCMC) methods. The Gibbs sampler is a well-known MCMC technique, widely applied in many signal processing problems. Drawing samples from univariate full-conditional distributions efficiently is essential for the practical application of the Gibbs sampler. In this work, we present a simple, self-tuned and extremely efficient MCMC algorithm which produces virtually independent samples from these univariate target densities. The proposal density used is self-tuned and tailored to the specific target, but it is not adaptive. Instead, the proposal is adjusted during an initial optimization stage, following a simple and extremely effective procedure. Hence, we have named the newly proposed approach as FUSS (Fast Universal Self-tuned Sampler), as it can be used to sample from any bounded univariate distribution and also from any bounded multi-variate distribution, either directly or by embedding it within a Gibbs sampler. Numerical experiments, on several synthetic data sets (including a challenging parameter estimation problem in a chaotic system) and a high-dimensional financial signal processing problem, show its good performance in terms of speed and estimation accuracy.

KW - Adaptive rejection Metropolis sampling

KW - Bayesian inference

KW - Gibbs sampling

KW - Markov Chain Monte Carlo (MCMC)

KW - Metropolis within Gibbs

UR - http://www.scopus.com/inward/record.url?scp=84928191257&partnerID=8YFLogxK

U2 - 10.1016/j.dsp.2015.04.005

DO - 10.1016/j.dsp.2015.04.005

M3 - Article

VL - 47

SP - 68

EP - 83

JO - Digital Signal Processing

JF - Digital Signal Processing

SN - 1051-2004

ER -