Rounding-based Moves for Metric Labeling

Abstract : Metric labeling is a special case of energy minimization for pairwise Markov random fields. The energy function consists of arbitrary unary potentials, and pairwise potentials that are proportional to a given metric distance function over the label set. Popular methods for solving metric labeling include (i) move-making algorithms, which iteratively solve a minimum st-cut problem; and (ii) the linear programming (LP) relaxation based approach. In order to convert the fractional solution of the LP relaxation to an integer solution, several randomized rounding procedures have been developed in the literature. We consider a large class of parallel rounding procedures, and design move-making algorithms that closely mimic them. We prove that the multiplicative bound of a move-making algorithm exactly matches the approximation factor of the corresponding rounding procedure for any arbitrary distance function. Our analysis includes all known results for move-making algorithms as special cases.
Type de document :
Communication dans un congrès
NIPS - Advances in Neural Information Processing Systems, 2014, Montreal, Canada. 2014
Liste complète des métadonnées

Littérature citée [25 références]  Voir  Masquer  Télécharger

https://hal.inria.fr/hal-01069910
Contributeur : M. Pawan Kumar <>
Soumis le : mardi 30 septembre 2014 - 10:57:46
Dernière modification le : lundi 1 octobre 2018 - 17:00:03
Document(s) archivé(s) le : mercredi 31 décembre 2014 - 10:35:47

Fichier

K-NIPS2014.pdf
Fichiers produits par l'(les) auteur(s)

Identifiants

  • HAL Id : hal-01069910, version 1

Collections

Citation

M. Pawan Kumar. Rounding-based Moves for Metric Labeling. NIPS - Advances in Neural Information Processing Systems, 2014, Montreal, Canada. 2014. 〈hal-01069910〉

Partager

Métriques

Consultations de la notice

171

Téléchargements de fichiers

168