Enregistré dans:
Détails bibliographiques
Auteurs principaux: Kemertas, Mete, Farahmand, Amir-massoud, Jepson, Allan D.
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:https://arxiv.org/abs/2504.02067
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866910902445408256
author Kemertas, Mete
Farahmand, Amir-massoud
Jepson, Allan D.
author_facet Kemertas, Mete
Farahmand, Amir-massoud
Jepson, Allan D.
contents Developing a contemporary optimal transport (OT) solver requires navigating trade-offs among several critical requirements: GPU parallelization, scalability to high-dimensional problems, theoretical convergence guarantees, empirical performance in terms of precision versus runtime, and numerical stability in practice. With these challenges in mind, we introduce a specialized truncated Newton algorithm for entropic-regularized OT. In addition to proving that locally quadratic convergence is possible without assuming a Lipschitz Hessian, we provide strategies to maximally exploit the high rate of local convergence in practice. Our GPU-parallel algorithm exhibits exceptionally favorable runtime performance, achieving high precision orders of magnitude faster than many existing alternatives. This is evidenced by wall-clock time experiments on 24 problem sets (12 datasets $\times$ 2 cost functions). The scalability of the algorithm is showcased on an extremely large OT problem with $n \approx 10^6$, solved approximately under weak entopric regularization.
format Preprint
id arxiv_https___arxiv_org_abs_2504_02067
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle A Truncated Newton Method for Optimal Transport
Kemertas, Mete
Farahmand, Amir-massoud
Jepson, Allan D.
Machine Learning
Mathematical Software
Optimization and Control
G.3; G.4; I.4.0
Developing a contemporary optimal transport (OT) solver requires navigating trade-offs among several critical requirements: GPU parallelization, scalability to high-dimensional problems, theoretical convergence guarantees, empirical performance in terms of precision versus runtime, and numerical stability in practice. With these challenges in mind, we introduce a specialized truncated Newton algorithm for entropic-regularized OT. In addition to proving that locally quadratic convergence is possible without assuming a Lipschitz Hessian, we provide strategies to maximally exploit the high rate of local convergence in practice. Our GPU-parallel algorithm exhibits exceptionally favorable runtime performance, achieving high precision orders of magnitude faster than many existing alternatives. This is evidenced by wall-clock time experiments on 24 problem sets (12 datasets $\times$ 2 cost functions). The scalability of the algorithm is showcased on an extremely large OT problem with $n \approx 10^6$, solved approximately under weak entopric regularization.
title A Truncated Newton Method for Optimal Transport
topic Machine Learning
Mathematical Software
Optimization and Control
G.3; G.4; I.4.0
url https://arxiv.org/abs/2504.02067