Smooth and strong

MAP inference with linear convergence

Ofer Meshi, Mehrdad Mahdavi, Alexander G. Schwing

Research output: Contribution to journalConference article

12 Citations (Scopus)

Abstract

Maximum a-posteriori (MAP) inference is an important task for many applications. Although the standard formulation gives rise to a hard combinatorial optimization problem, several effective approximations have been proposed and studied in recent years. We focus on linear programming (LP) relaxations, which have achieved state-of-the-art performance in many applications. However, optimization of the resulting program is in general challenging due to non-smoothness and complex non-separable constraints. Therefore, in this work we study the benefits of augmenting the objective function of the relaxation with strong convexity. Specifically, we introduce strong convexity by adding a quadratic term to the LP relaxation objective. We provide theoretical guarantees for the resulting programs, bounding the difference between their optimal value and the original optimum. Further, we propose suitable optimization algorithms and analyze their convergence.

Original languageEnglish (US)
Pages (from-to)298-306
Number of pages9
JournalAdvances in Neural Information Processing Systems
Volume2015-January
StatePublished - Jan 1 2015
Event29th Annual Conference on Neural Information Processing Systems, NIPS 2015 - Montreal, Canada
Duration: Dec 7 2015Dec 12 2015

Fingerprint

Linear programming
Combinatorial optimization

All Science Journal Classification (ASJC) codes

  • Computer Networks and Communications
  • Information Systems
  • Signal Processing

Cite this

Meshi, Ofer ; Mahdavi, Mehrdad ; Schwing, Alexander G. / Smooth and strong : MAP inference with linear convergence. In: Advances in Neural Information Processing Systems. 2015 ; Vol. 2015-January. pp. 298-306.
@article{6a8a6d4fcc244a258264b44c5eda3875,
title = "Smooth and strong: MAP inference with linear convergence",
abstract = "Maximum a-posteriori (MAP) inference is an important task for many applications. Although the standard formulation gives rise to a hard combinatorial optimization problem, several effective approximations have been proposed and studied in recent years. We focus on linear programming (LP) relaxations, which have achieved state-of-the-art performance in many applications. However, optimization of the resulting program is in general challenging due to non-smoothness and complex non-separable constraints. Therefore, in this work we study the benefits of augmenting the objective function of the relaxation with strong convexity. Specifically, we introduce strong convexity by adding a quadratic term to the LP relaxation objective. We provide theoretical guarantees for the resulting programs, bounding the difference between their optimal value and the original optimum. Further, we propose suitable optimization algorithms and analyze their convergence.",
author = "Ofer Meshi and Mehrdad Mahdavi and Schwing, {Alexander G.}",
year = "2015",
month = "1",
day = "1",
language = "English (US)",
volume = "2015-January",
pages = "298--306",
journal = "Advances in Neural Information Processing Systems",
issn = "1049-5258",

}

Smooth and strong : MAP inference with linear convergence. / Meshi, Ofer; Mahdavi, Mehrdad; Schwing, Alexander G.

In: Advances in Neural Information Processing Systems, Vol. 2015-January, 01.01.2015, p. 298-306.

Research output: Contribution to journalConference article

TY - JOUR

T1 - Smooth and strong

T2 - MAP inference with linear convergence

AU - Meshi, Ofer

AU - Mahdavi, Mehrdad

AU - Schwing, Alexander G.

PY - 2015/1/1

Y1 - 2015/1/1

N2 - Maximum a-posteriori (MAP) inference is an important task for many applications. Although the standard formulation gives rise to a hard combinatorial optimization problem, several effective approximations have been proposed and studied in recent years. We focus on linear programming (LP) relaxations, which have achieved state-of-the-art performance in many applications. However, optimization of the resulting program is in general challenging due to non-smoothness and complex non-separable constraints. Therefore, in this work we study the benefits of augmenting the objective function of the relaxation with strong convexity. Specifically, we introduce strong convexity by adding a quadratic term to the LP relaxation objective. We provide theoretical guarantees for the resulting programs, bounding the difference between their optimal value and the original optimum. Further, we propose suitable optimization algorithms and analyze their convergence.

AB - Maximum a-posteriori (MAP) inference is an important task for many applications. Although the standard formulation gives rise to a hard combinatorial optimization problem, several effective approximations have been proposed and studied in recent years. We focus on linear programming (LP) relaxations, which have achieved state-of-the-art performance in many applications. However, optimization of the resulting program is in general challenging due to non-smoothness and complex non-separable constraints. Therefore, in this work we study the benefits of augmenting the objective function of the relaxation with strong convexity. Specifically, we introduce strong convexity by adding a quadratic term to the LP relaxation objective. We provide theoretical guarantees for the resulting programs, bounding the difference between their optimal value and the original optimum. Further, we propose suitable optimization algorithms and analyze their convergence.

UR - http://www.scopus.com/inward/record.url?scp=84965175226&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84965175226&partnerID=8YFLogxK

M3 - Conference article

VL - 2015-January

SP - 298

EP - 306

JO - Advances in Neural Information Processing Systems

JF - Advances in Neural Information Processing Systems

SN - 1049-5258

ER -