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. 2307.10352
15
10

Properties of Discrete Sliced Wasserstein Losses

19 July 2023
Eloi Tanguy
Rémi Flamary
J. Delon
ArXivPDFHTML
Abstract

The Sliced Wasserstein (SW) distance has become a popular alternative to the Wasserstein distance for comparing probability measures. Widespread applications include image processing, domain adaptation and generative modelling, where it is common to optimise some parameters in order to minimise SW, which serves as a loss function between discrete probability measures (since measures admitting densities are numerically unattainable). All these optimisation problems bear the same sub-problem, which is minimising the Sliced Wasserstein energy. In this paper we study the properties of E:Y⟼SW22(γY,γZ)\mathcal{E}: Y \longmapsto \mathrm{SW}_2^2(\gamma_Y, \gamma_Z)E:Y⟼SW22​(γY​,γZ​), i.e. the SW distance between two uniform discrete measures with the same amount of points as a function of the support Y∈Rn×dY \in \mathbb{R}^{n \times d}Y∈Rn×d of one of the measures. We investigate the regularity and optimisation properties of this energy, as well as its Monte-Carlo approximation Ep\mathcal{E}_pEp​ (estimating the expectation in SW using only ppp samples) and show convergence results on the critical points of Ep\mathcal{E}_pEp​ to those of E\mathcal{E}E, as well as an almost-sure uniform convergence and a uniform Central Limit result on the process Ep(Y)\mathcal{E}_p(Y)Ep​(Y). Finally, we show that in a certain sense, Stochastic Gradient Descent methods minimising E\mathcal{E}E and Ep\mathcal{E}_pEp​ converge towards (Clarke) critical points of these energies.

View on arXiv
Comments on this paper