Lifted tree-reweighted variational inference

Hung Hai Bui, Tuyen N. Huynh, David Sontag

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

We analyze variational inference for highly symmetric graphical models such as those arising from first-order probabilistic models. We first show that for these graphical models, the treereweighted variational objective lends itself to a compact lifted formulation which can be solved much more efficiently than the standard TRW formulation for the ground graphical model. Compared to earlier work on lifted belief propagation, our formulation leads to a convex optimization problem for lifted marginal inference and provides an upper bound on the partition function. We provide two approaches for improving the lifted TRW upper bound. The first is a method for efficiently computing maximum spanning trees in highly symmetric graphs, which can be used to optimize the TRW edge appearance probabilities. The second is a method for tightening the relaxation of the marginal polytope using lifted cycle inequalities and novel exchangeable cluster consistency constraints.

Original languageEnglish (US)
Title of host publicationUncertainty in Artificial Intelligence - Proceedings of the 30th Conference, UAI 2014
PublisherAUAI Press
Pages92-101
Number of pages10
ISBN (Print)9780974903910
StatePublished - 2014
Event30th Conference on Uncertainty in Artificial Intelligence, UAI 2014 - Quebec City, Canada
Duration: Jul 23 2014Jul 27 2014

Other

Other30th Conference on Uncertainty in Artificial Intelligence, UAI 2014
CountryCanada
CityQuebec City
Period7/23/147/27/14

Fingerprint

Convex optimization
Statistical Models

ASJC Scopus subject areas

  • Artificial Intelligence

Cite this

Bui, H. H., Huynh, T. N., & Sontag, D. (2014). Lifted tree-reweighted variational inference. In Uncertainty in Artificial Intelligence - Proceedings of the 30th Conference, UAI 2014 (pp. 92-101). AUAI Press.

Lifted tree-reweighted variational inference. / Bui, Hung Hai; Huynh, Tuyen N.; Sontag, David.

Uncertainty in Artificial Intelligence - Proceedings of the 30th Conference, UAI 2014. AUAI Press, 2014. p. 92-101.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Bui, HH, Huynh, TN & Sontag, D 2014, Lifted tree-reweighted variational inference. in Uncertainty in Artificial Intelligence - Proceedings of the 30th Conference, UAI 2014. AUAI Press, pp. 92-101, 30th Conference on Uncertainty in Artificial Intelligence, UAI 2014, Quebec City, Canada, 7/23/14.
Bui HH, Huynh TN, Sontag D. Lifted tree-reweighted variational inference. In Uncertainty in Artificial Intelligence - Proceedings of the 30th Conference, UAI 2014. AUAI Press. 2014. p. 92-101
Bui, Hung Hai ; Huynh, Tuyen N. ; Sontag, David. / Lifted tree-reweighted variational inference. Uncertainty in Artificial Intelligence - Proceedings of the 30th Conference, UAI 2014. AUAI Press, 2014. pp. 92-101
@inproceedings{a5a41d880fe346d2a3d40487d2d7d9b2,
title = "Lifted tree-reweighted variational inference",
abstract = "We analyze variational inference for highly symmetric graphical models such as those arising from first-order probabilistic models. We first show that for these graphical models, the treereweighted variational objective lends itself to a compact lifted formulation which can be solved much more efficiently than the standard TRW formulation for the ground graphical model. Compared to earlier work on lifted belief propagation, our formulation leads to a convex optimization problem for lifted marginal inference and provides an upper bound on the partition function. We provide two approaches for improving the lifted TRW upper bound. The first is a method for efficiently computing maximum spanning trees in highly symmetric graphs, which can be used to optimize the TRW edge appearance probabilities. The second is a method for tightening the relaxation of the marginal polytope using lifted cycle inequalities and novel exchangeable cluster consistency constraints.",
author = "Bui, {Hung Hai} and Huynh, {Tuyen N.} and David Sontag",
year = "2014",
language = "English (US)",
isbn = "9780974903910",
pages = "92--101",
booktitle = "Uncertainty in Artificial Intelligence - Proceedings of the 30th Conference, UAI 2014",
publisher = "AUAI Press",

}

TY - GEN

T1 - Lifted tree-reweighted variational inference

AU - Bui, Hung Hai

AU - Huynh, Tuyen N.

AU - Sontag, David

PY - 2014

Y1 - 2014

N2 - We analyze variational inference for highly symmetric graphical models such as those arising from first-order probabilistic models. We first show that for these graphical models, the treereweighted variational objective lends itself to a compact lifted formulation which can be solved much more efficiently than the standard TRW formulation for the ground graphical model. Compared to earlier work on lifted belief propagation, our formulation leads to a convex optimization problem for lifted marginal inference and provides an upper bound on the partition function. We provide two approaches for improving the lifted TRW upper bound. The first is a method for efficiently computing maximum spanning trees in highly symmetric graphs, which can be used to optimize the TRW edge appearance probabilities. The second is a method for tightening the relaxation of the marginal polytope using lifted cycle inequalities and novel exchangeable cluster consistency constraints.

AB - We analyze variational inference for highly symmetric graphical models such as those arising from first-order probabilistic models. We first show that for these graphical models, the treereweighted variational objective lends itself to a compact lifted formulation which can be solved much more efficiently than the standard TRW formulation for the ground graphical model. Compared to earlier work on lifted belief propagation, our formulation leads to a convex optimization problem for lifted marginal inference and provides an upper bound on the partition function. We provide two approaches for improving the lifted TRW upper bound. The first is a method for efficiently computing maximum spanning trees in highly symmetric graphs, which can be used to optimize the TRW edge appearance probabilities. The second is a method for tightening the relaxation of the marginal polytope using lifted cycle inequalities and novel exchangeable cluster consistency constraints.

UR - http://www.scopus.com/inward/record.url?scp=84923287582&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84923287582&partnerID=8YFLogxK

M3 - Conference contribution

SN - 9780974903910

SP - 92

EP - 101

BT - Uncertainty in Artificial Intelligence - Proceedings of the 30th Conference, UAI 2014

PB - AUAI Press

ER -