Barrier Frank-Wolfe for marginal inference

Rahul G. Krishnan, Simon Lacoste-Julien, David Sontag

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

Abstract

We introduce a globally-convergent algorithm for optimizing the tree-reweighted (TRW) variational objective over the marginal polytope. The algorithm is based on the conditional gradient method (Frank-Wolfe) and moves pseudomarginals within the marginal polytope through repeated maximum a posteriori (MAP) calls. This modular structure enables us to leverage black-box MAP solvers (both exact and approximate) for variational inference, and obtains more accurate results than tree-reweighted algorithms that optimize over the local consistency relaxation. Theoretically, we bound the sub-optimality for the proposed algorithm despite the TRW objective having unbounded gradients at the boundary of the marginal polytope. Empirically, we demonstrate the increased quality of results found by tightening the relaxation over the marginal polytope as well as the spanning tree polytope on synthetic and real-world instances.

Original languageEnglish (US)
Title of host publicationAdvances in Neural Information Processing Systems
PublisherNeural information processing systems foundation
Pages532-540
Number of pages9
Volume2015-January
StatePublished - 2015
Event29th Annual Conference on Neural Information Processing Systems, NIPS 2015 - Montreal, Canada
Duration: Dec 7 2015Dec 12 2015

Other

Other29th Annual Conference on Neural Information Processing Systems, NIPS 2015
CountryCanada
CityMontreal
Period12/7/1512/12/15

Fingerprint

Gradient methods

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Information Systems
  • Signal Processing

Cite this

Krishnan, R. G., Lacoste-Julien, S., & Sontag, D. (2015). Barrier Frank-Wolfe for marginal inference. In Advances in Neural Information Processing Systems (Vol. 2015-January, pp. 532-540). Neural information processing systems foundation.

Barrier Frank-Wolfe for marginal inference. / Krishnan, Rahul G.; Lacoste-Julien, Simon; Sontag, David.

Advances in Neural Information Processing Systems. Vol. 2015-January Neural information processing systems foundation, 2015. p. 532-540.

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

Krishnan, RG, Lacoste-Julien, S & Sontag, D 2015, Barrier Frank-Wolfe for marginal inference. in Advances in Neural Information Processing Systems. vol. 2015-January, Neural information processing systems foundation, pp. 532-540, 29th Annual Conference on Neural Information Processing Systems, NIPS 2015, Montreal, Canada, 12/7/15.
Krishnan RG, Lacoste-Julien S, Sontag D. Barrier Frank-Wolfe for marginal inference. In Advances in Neural Information Processing Systems. Vol. 2015-January. Neural information processing systems foundation. 2015. p. 532-540
Krishnan, Rahul G. ; Lacoste-Julien, Simon ; Sontag, David. / Barrier Frank-Wolfe for marginal inference. Advances in Neural Information Processing Systems. Vol. 2015-January Neural information processing systems foundation, 2015. pp. 532-540
@inproceedings{a17834fd3caf465586525a1ac9e399a5,
title = "Barrier Frank-Wolfe for marginal inference",
abstract = "We introduce a globally-convergent algorithm for optimizing the tree-reweighted (TRW) variational objective over the marginal polytope. The algorithm is based on the conditional gradient method (Frank-Wolfe) and moves pseudomarginals within the marginal polytope through repeated maximum a posteriori (MAP) calls. This modular structure enables us to leverage black-box MAP solvers (both exact and approximate) for variational inference, and obtains more accurate results than tree-reweighted algorithms that optimize over the local consistency relaxation. Theoretically, we bound the sub-optimality for the proposed algorithm despite the TRW objective having unbounded gradients at the boundary of the marginal polytope. Empirically, we demonstrate the increased quality of results found by tightening the relaxation over the marginal polytope as well as the spanning tree polytope on synthetic and real-world instances.",
author = "Krishnan, {Rahul G.} and Simon Lacoste-Julien and David Sontag",
year = "2015",
language = "English (US)",
volume = "2015-January",
pages = "532--540",
booktitle = "Advances in Neural Information Processing Systems",
publisher = "Neural information processing systems foundation",

}

TY - GEN

T1 - Barrier Frank-Wolfe for marginal inference

AU - Krishnan, Rahul G.

AU - Lacoste-Julien, Simon

AU - Sontag, David

PY - 2015

Y1 - 2015

N2 - We introduce a globally-convergent algorithm for optimizing the tree-reweighted (TRW) variational objective over the marginal polytope. The algorithm is based on the conditional gradient method (Frank-Wolfe) and moves pseudomarginals within the marginal polytope through repeated maximum a posteriori (MAP) calls. This modular structure enables us to leverage black-box MAP solvers (both exact and approximate) for variational inference, and obtains more accurate results than tree-reweighted algorithms that optimize over the local consistency relaxation. Theoretically, we bound the sub-optimality for the proposed algorithm despite the TRW objective having unbounded gradients at the boundary of the marginal polytope. Empirically, we demonstrate the increased quality of results found by tightening the relaxation over the marginal polytope as well as the spanning tree polytope on synthetic and real-world instances.

AB - We introduce a globally-convergent algorithm for optimizing the tree-reweighted (TRW) variational objective over the marginal polytope. The algorithm is based on the conditional gradient method (Frank-Wolfe) and moves pseudomarginals within the marginal polytope through repeated maximum a posteriori (MAP) calls. This modular structure enables us to leverage black-box MAP solvers (both exact and approximate) for variational inference, and obtains more accurate results than tree-reweighted algorithms that optimize over the local consistency relaxation. Theoretically, we bound the sub-optimality for the proposed algorithm despite the TRW objective having unbounded gradients at the boundary of the marginal polytope. Empirically, we demonstrate the increased quality of results found by tightening the relaxation over the marginal polytope as well as the spanning tree polytope on synthetic and real-world instances.

UR - http://www.scopus.com/inward/record.url?scp=84965120278&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84965120278&partnerID=8YFLogxK

M3 - Conference contribution

VL - 2015-January

SP - 532

EP - 540

BT - Advances in Neural Information Processing Systems

PB - Neural information processing systems foundation

ER -