ResearchTrend.AI
  • Papers
  • Communities
  • Events
  • Blog
  • Pricing
Papers
Communities
Social Events
Terms and Conditions
Pricing
Parameter LabParameter LabTwitterGitHubLinkedInBlueskyYoutube

© 2025 ResearchTrend.AI, All rights reserved.

  1. Home
  2. Papers
  3. 2311.01404
30
3

Normalizing flows as approximations of optimal transport maps via linear-control neural ODEs

2 November 2023
A. Scagliotti
Sara Farinelli
ArXivPDFHTML
Abstract

The term "Normalizing Flows" is related to the task of constructing invertible transport maps between probability measures by means of deep neural networks. In this paper, we consider the problem of recovering the W2W_2W2​-optimal transport map TTT between absolutely continuous measures μ,ν∈P(Rn)\mu,\nu\in\mathcal{P}(\mathbb{R}^n)μ,ν∈P(Rn) as the flow of a linear-control neural ODE. We first show that, under suitable assumptions on μ,ν\mu,\nuμ,ν and on the controlled vector fields, the optimal transport map is contained in the Cc0C^0_cCc0​-closure of the flows generated by the system. Assuming that discrete approximations μN,νN\mu_N,\nu_NμN​,νN​ of the original measures μ,ν\mu,\nuμ,ν are available, we use a discrete optimal coupling γN\gamma_NγN​ to define an optimal control problem. With a Γ\GammaΓ-convergence argument, we prove that its solutions correspond to flows that approximate the optimal transport map TTT. Finally, taking advantage of the Pontryagin Maximum Principle, we propose an iterative numerical scheme for the resolution of the optimal control problem, resulting in an algorithm for the practical computation of the approximated optimal transport map.

View on arXiv
Comments on this paper