Semidefinite Relaxations of the Gromov-Wasserstein Distance

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Chen, Junyu, Nguyen, Binh T., Koh, Shang Hui, Soh, Yong Sheng
Format: Preprint
Publié: 2023
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866913547849564160
author Chen, Junyu
Nguyen, Binh T.
Koh, Shang Hui
Soh, Yong Sheng
author_facet Chen, Junyu
Nguyen, Binh T.
Koh, Shang Hui
Soh, Yong Sheng
contents The Gromov-Wasserstein (GW) distance is an extension of the optimal transport problem that allows one to match objects between incomparable spaces. At its core, the GW distance is specified as the solution of a non-convex quadratic program and is not known to be tractable to solve. In particular, existing solvers for the GW distance are only able to find locally optimal solutions. In this work, we propose a semi-definite programming (SDP) relaxation of the GW distance. The relaxation can be viewed as the Lagrangian dual of the GW distance augmented with constraints that relate to the linear and quadratic terms of transportation plans. In particular, our relaxation provides a tractable (polynomial-time) algorithm to compute globally optimal transportation plans (in some instances) together with an accompanying proof of global optimality. Our numerical experiments suggest that the proposed relaxation is strong in that it frequently computes the globally optimal solution. Our Python implementation is available at https://github.com/tbng/gwsdp.
format Preprint
id arxiv_https___arxiv_org_abs_2312_14572
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Semidefinite Relaxations of the Gromov-Wasserstein Distance
Chen, Junyu
Nguyen, Binh T.
Koh, Shang Hui
Soh, Yong Sheng
Optimization and Control
Machine Learning
The Gromov-Wasserstein (GW) distance is an extension of the optimal transport problem that allows one to match objects between incomparable spaces. At its core, the GW distance is specified as the solution of a non-convex quadratic program and is not known to be tractable to solve. In particular, existing solvers for the GW distance are only able to find locally optimal solutions. In this work, we propose a semi-definite programming (SDP) relaxation of the GW distance. The relaxation can be viewed as the Lagrangian dual of the GW distance augmented with constraints that relate to the linear and quadratic terms of transportation plans. In particular, our relaxation provides a tractable (polynomial-time) algorithm to compute globally optimal transportation plans (in some instances) together with an accompanying proof of global optimality. Our numerical experiments suggest that the proposed relaxation is strong in that it frequently computes the globally optimal solution. Our Python implementation is available at https://github.com/tbng/gwsdp.
title Semidefinite Relaxations of the Gromov-Wasserstein Distance
topic Optimization and Control
Machine Learning
url https://arxiv.org/abs/2312.14572