Sampling from Gaussian Markov random fields using stationary and non-stationary subgraph perturbations

Ying Liu, Oliver Kosut, Alan S. Willsky

Research output: Contribution to journalArticle

4 Citations (Scopus)

Abstract

Gaussian Markov random fields (GMRFs) or Gaussian graphical models have been widely used in many applications. Efficiently drawing samples from GMRFs has been an important research problem. In this paper, we introduce the subgraph perturbation sampling algorithm, which makes use of any pre-existing tractable inference algorithm for a subgraph by perturbing this algorithm so as to yield asymptotically exact samples for the intended distribution. We study the stationary version where a single fixed subgraph is used in all iterations, as well as the non-stationary version where tractable subgraphs are adaptively selected. The subgraphs used can have any structure for which efficient inference algorithms exist: for example, tree-structured, low tree-width, or having a small feedback vertex set. We present new theoretical results that give convergence guarantees for both stationary and non-stationary graphical splittings. Our experiments using both simulated models and large-scale real models demonstrate that this subgraph perturbation algorithm efficiently yields accurate samples for many graph topologies.

Original languageEnglish (US)
Article number6967838
Pages (from-to)576-589
Number of pages14
JournalIEEE Transactions on Signal Processing
Volume63
Issue number3
DOIs
StatePublished - Feb 1 2015

Fingerprint

Sampling
Topology
Feedback
Experiments

Keywords

  • Feedback vertex set
  • Gaussian graphical models
  • Gaussian Markov random fields
  • graphical splittings

ASJC Scopus subject areas

  • Electrical and Electronic Engineering
  • Signal Processing

Cite this

Sampling from Gaussian Markov random fields using stationary and non-stationary subgraph perturbations. / Liu, Ying; Kosut, Oliver; Willsky, Alan S.

In: IEEE Transactions on Signal Processing, Vol. 63, No. 3, 6967838, 01.02.2015, p. 576-589.

Research output: Contribution to journalArticle

@article{d7324787a5324dc595ba290691d030c3,
title = "Sampling from Gaussian Markov random fields using stationary and non-stationary subgraph perturbations",
abstract = "Gaussian Markov random fields (GMRFs) or Gaussian graphical models have been widely used in many applications. Efficiently drawing samples from GMRFs has been an important research problem. In this paper, we introduce the subgraph perturbation sampling algorithm, which makes use of any pre-existing tractable inference algorithm for a subgraph by perturbing this algorithm so as to yield asymptotically exact samples for the intended distribution. We study the stationary version where a single fixed subgraph is used in all iterations, as well as the non-stationary version where tractable subgraphs are adaptively selected. The subgraphs used can have any structure for which efficient inference algorithms exist: for example, tree-structured, low tree-width, or having a small feedback vertex set. We present new theoretical results that give convergence guarantees for both stationary and non-stationary graphical splittings. Our experiments using both simulated models and large-scale real models demonstrate that this subgraph perturbation algorithm efficiently yields accurate samples for many graph topologies.",
keywords = "Feedback vertex set, Gaussian graphical models, Gaussian Markov random fields, graphical splittings",
author = "Ying Liu and Oliver Kosut and Willsky, {Alan S.}",
year = "2015",
month = "2",
day = "1",
doi = "10.1109/TSP.2014.2375134",
language = "English (US)",
volume = "63",
pages = "576--589",
journal = "IEEE Transactions on Signal Processing",
issn = "1053-587X",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "3",

}

TY - JOUR

T1 - Sampling from Gaussian Markov random fields using stationary and non-stationary subgraph perturbations

AU - Liu, Ying

AU - Kosut, Oliver

AU - Willsky, Alan S.

PY - 2015/2/1

Y1 - 2015/2/1

N2 - Gaussian Markov random fields (GMRFs) or Gaussian graphical models have been widely used in many applications. Efficiently drawing samples from GMRFs has been an important research problem. In this paper, we introduce the subgraph perturbation sampling algorithm, which makes use of any pre-existing tractable inference algorithm for a subgraph by perturbing this algorithm so as to yield asymptotically exact samples for the intended distribution. We study the stationary version where a single fixed subgraph is used in all iterations, as well as the non-stationary version where tractable subgraphs are adaptively selected. The subgraphs used can have any structure for which efficient inference algorithms exist: for example, tree-structured, low tree-width, or having a small feedback vertex set. We present new theoretical results that give convergence guarantees for both stationary and non-stationary graphical splittings. Our experiments using both simulated models and large-scale real models demonstrate that this subgraph perturbation algorithm efficiently yields accurate samples for many graph topologies.

AB - Gaussian Markov random fields (GMRFs) or Gaussian graphical models have been widely used in many applications. Efficiently drawing samples from GMRFs has been an important research problem. In this paper, we introduce the subgraph perturbation sampling algorithm, which makes use of any pre-existing tractable inference algorithm for a subgraph by perturbing this algorithm so as to yield asymptotically exact samples for the intended distribution. We study the stationary version where a single fixed subgraph is used in all iterations, as well as the non-stationary version where tractable subgraphs are adaptively selected. The subgraphs used can have any structure for which efficient inference algorithms exist: for example, tree-structured, low tree-width, or having a small feedback vertex set. We present new theoretical results that give convergence guarantees for both stationary and non-stationary graphical splittings. Our experiments using both simulated models and large-scale real models demonstrate that this subgraph perturbation algorithm efficiently yields accurate samples for many graph topologies.

KW - Feedback vertex set

KW - Gaussian graphical models

KW - Gaussian Markov random fields

KW - graphical splittings

UR - http://www.scopus.com/inward/record.url?scp=84920114730&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84920114730&partnerID=8YFLogxK

U2 - 10.1109/TSP.2014.2375134

DO - 10.1109/TSP.2014.2375134

M3 - Article

AN - SCOPUS:84920114730

VL - 63

SP - 576

EP - 589

JO - IEEE Transactions on Signal Processing

JF - IEEE Transactions on Signal Processing

SN - 1053-587X

IS - 3

M1 - 6967838

ER -