Overrelaxed Sinkhorn-Knopp Algorithm for Regularized Optimal Transport
dc.contributor.author | THIBAULT, Alexis | |
hal.structure.identifier | CEntre de REcherches en MAthématiques de la DEcision [CEREMADE] | |
dc.contributor.author | CHIZAT, Lenaic | |
hal.structure.identifier | Institut National des Sciences Appliquées - Toulouse [INSA Toulouse] | |
hal.structure.identifier | Institut de Mathématiques de Toulouse UMR5219 [IMT] | |
dc.contributor.author | DOSSAL, Charles | |
hal.structure.identifier | Institut de Mathématiques de Bordeaux [IMB] | |
dc.contributor.author | PAPADAKIS, Nicolas | |
dc.date.accessioned | 2024-04-04T03:00:18Z | |
dc.date.available | 2024-04-04T03:00:18Z | |
dc.date.conference | 2017-12-09 | |
dc.identifier.uri | https://oskar-bordeaux.fr/handle/20.500.12278/192806 | |
dc.description.abstractEn | This article describes a method for quickly computing the solution to the regularized optimal transport problem. It generalizes and improves upon the widely-used iterative Bregman projections algorithm (or Sinkhorn-Knopp algorithm). The idea is to overrelax the Bregman projection operators, allowing for faster convergence. In practice this corresponds to elevating the diagonal scaling factors to a given power, at each step of the algorithm. We propose a simple method for establishing global convergence by ensuring the decrease of a Lyapunov function at each step. An adaptive choice of overrelaxation parameter based on the Lyapunov function is constructed. We also suggest a heuristic to choose a suitable asymptotic overrelaxation parameter, based on a local convergence analysis. Our numerical experiments show a gain in convergence speed by an order of magnitude in certain regimes. | |
dc.description.sponsorship | Generalized Optimal Transport Models for Image processing - ANR-16-CE33-0010 | |
dc.language.iso | en | |
dc.title.en | Overrelaxed Sinkhorn-Knopp Algorithm for Regularized Optimal Transport | |
dc.type | Communication dans un congrès | |
dc.subject.hal | Informatique [cs]/Traitement du signal et de l'image | |
dc.subject.hal | Mathématiques [math]/Optimisation et contrôle [math.OC] | |
dc.identifier.arxiv | 1711.01851 | |
bordeaux.hal.laboratories | Institut de Mathématiques de Bordeaux (IMB) - UMR 5251 | * |
bordeaux.institution | Université de Bordeaux | |
bordeaux.institution | Bordeaux INP | |
bordeaux.institution | CNRS | |
bordeaux.conference.title | NIPS Workshop on Optimal Transport & Machine Learning (OTML'17) | |
bordeaux.country | US | |
bordeaux.conference.city | Long Beach | |
bordeaux.peerReviewed | oui | |
hal.identifier | hal-01629985 | |
hal.version | 1 | |
hal.invited | non | |
hal.proceedings | oui | |
hal.popular | non | |
hal.audience | Internationale | |
hal.origin.link | https://hal.archives-ouvertes.fr//hal-01629985v1 | |
bordeaux.COinS | ctx_ver=Z39.88-2004&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.au=THIBAULT,%20Alexis&CHIZAT,%20Lenaic&DOSSAL,%20Charles&PAPADAKIS,%20Nicolas&rft.genre=unknown |
Fichier(s) constituant ce document
Fichiers | Taille | Format | Vue |
---|---|---|---|
Il n'y a pas de fichiers associés à ce document. |