Constrained likelihood for reconstructing a directed acyclic Gaussian graph

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

Directed acyclic graphs are widely used to describe directional pairwise relations. Such relations are estimated by reconstructing a directed acyclic graph's structure, which is challenging when the ordering of nodes of the graph is unknown. In such a situation, existing methods such as the neighbourhood and search-and-score methods have high estimation errors or computational complexities, especially when a local or sequential approach is used to enumerate edge directions by testing or optimizing a criterion locally, as a local method may break down even for moderately sized graphs. We propose a novel approach to simultaneously identifying all estimable directed edges and model parameters, using constrained maximum likelihood with nonconvex constraints. We develop a constraint reduction method that constructs a set of active constraints from super-exponentially many constraints. This, coupled with an alternating direction method of multipliers and a difference convex method, permits efficient computation for large-graph learning. We show that the proposed method consistently reconstructs identifiable directions of the true graph and achieves the optimal performance in terms of parameter estimation. Numerically, the method compares favourably with competitors. A protein network is analysed to demonstrate that the proposed method can make a difference in identifying the network's structure.

Original languageEnglish (US)
Pages (from-to)109-125
Number of pages17
JournalBiometrika
Volume106
Issue number1
DOIs
StatePublished - Mar 1 2019

Fingerprint

Parameter estimation
Error analysis
Maximum likelihood
Computational complexity
Likelihood
Proteins
Testing
Graph in graph theory
Directed Acyclic Graph
methodology
Method of multipliers
Alternating Direction Method
Estimation Error
Graph
Reduction Method
multipliers
Network Structure
Maximum Likelihood
Breakdown
Parameter Estimation

Keywords

  • Acyclicity
  • Constraint reduction
  • Directional relation
  • Nonconvex minimization

PubMed: MeSH publication types

  • Journal Article

Cite this

Constrained likelihood for reconstructing a directed acyclic Gaussian graph. / Yuan, Yiping; Shen, Xiaotong T; Pan, Wei; Wang, Zizhuo.

In: Biometrika, Vol. 106, No. 1, 01.03.2019, p. 109-125.

Research output: Contribution to journalArticle

@article{c7bc0275820948fa934c0d6fa61baa58,
title = "Constrained likelihood for reconstructing a directed acyclic Gaussian graph",
abstract = "Directed acyclic graphs are widely used to describe directional pairwise relations. Such relations are estimated by reconstructing a directed acyclic graph's structure, which is challenging when the ordering of nodes of the graph is unknown. In such a situation, existing methods such as the neighbourhood and search-and-score methods have high estimation errors or computational complexities, especially when a local or sequential approach is used to enumerate edge directions by testing or optimizing a criterion locally, as a local method may break down even for moderately sized graphs. We propose a novel approach to simultaneously identifying all estimable directed edges and model parameters, using constrained maximum likelihood with nonconvex constraints. We develop a constraint reduction method that constructs a set of active constraints from super-exponentially many constraints. This, coupled with an alternating direction method of multipliers and a difference convex method, permits efficient computation for large-graph learning. We show that the proposed method consistently reconstructs identifiable directions of the true graph and achieves the optimal performance in terms of parameter estimation. Numerically, the method compares favourably with competitors. A protein network is analysed to demonstrate that the proposed method can make a difference in identifying the network's structure.",
keywords = "Acyclicity, Constraint reduction, Directional relation, Nonconvex minimization",
author = "Yiping Yuan and Shen, {Xiaotong T} and Wei Pan and Zizhuo Wang",
year = "2019",
month = "3",
day = "1",
doi = "10.1093/biomet/asy057",
language = "English (US)",
volume = "106",
pages = "109--125",
journal = "Biometrika",
issn = "0006-3444",
publisher = "Oxford University Press",
number = "1",

}

TY - JOUR

T1 - Constrained likelihood for reconstructing a directed acyclic Gaussian graph

AU - Yuan, Yiping

AU - Shen, Xiaotong T

AU - Pan, Wei

AU - Wang, Zizhuo

PY - 2019/3/1

Y1 - 2019/3/1

N2 - Directed acyclic graphs are widely used to describe directional pairwise relations. Such relations are estimated by reconstructing a directed acyclic graph's structure, which is challenging when the ordering of nodes of the graph is unknown. In such a situation, existing methods such as the neighbourhood and search-and-score methods have high estimation errors or computational complexities, especially when a local or sequential approach is used to enumerate edge directions by testing or optimizing a criterion locally, as a local method may break down even for moderately sized graphs. We propose a novel approach to simultaneously identifying all estimable directed edges and model parameters, using constrained maximum likelihood with nonconvex constraints. We develop a constraint reduction method that constructs a set of active constraints from super-exponentially many constraints. This, coupled with an alternating direction method of multipliers and a difference convex method, permits efficient computation for large-graph learning. We show that the proposed method consistently reconstructs identifiable directions of the true graph and achieves the optimal performance in terms of parameter estimation. Numerically, the method compares favourably with competitors. A protein network is analysed to demonstrate that the proposed method can make a difference in identifying the network's structure.

AB - Directed acyclic graphs are widely used to describe directional pairwise relations. Such relations are estimated by reconstructing a directed acyclic graph's structure, which is challenging when the ordering of nodes of the graph is unknown. In such a situation, existing methods such as the neighbourhood and search-and-score methods have high estimation errors or computational complexities, especially when a local or sequential approach is used to enumerate edge directions by testing or optimizing a criterion locally, as a local method may break down even for moderately sized graphs. We propose a novel approach to simultaneously identifying all estimable directed edges and model parameters, using constrained maximum likelihood with nonconvex constraints. We develop a constraint reduction method that constructs a set of active constraints from super-exponentially many constraints. This, coupled with an alternating direction method of multipliers and a difference convex method, permits efficient computation for large-graph learning. We show that the proposed method consistently reconstructs identifiable directions of the true graph and achieves the optimal performance in terms of parameter estimation. Numerically, the method compares favourably with competitors. A protein network is analysed to demonstrate that the proposed method can make a difference in identifying the network's structure.

KW - Acyclicity

KW - Constraint reduction

KW - Directional relation

KW - Nonconvex minimization

UR - http://www.scopus.com/inward/record.url?scp=85063230086&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85063230086&partnerID=8YFLogxK

U2 - 10.1093/biomet/asy057

DO - 10.1093/biomet/asy057

M3 - Article

C2 - 30799877

AN - SCOPUS:85063230086

VL - 106

SP - 109

EP - 125

JO - Biometrika

JF - Biometrika

SN - 0006-3444

IS - 1

ER -