Minimax-Optimal Two-Sample Test with Sliced Wasserstein

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Tran, Binh Thuan, Schreuder, Nicolas
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908622752055296
author Tran, Binh Thuan
Schreuder, Nicolas
author_facet Tran, Binh Thuan
Schreuder, Nicolas
contents We study the problem of nonparametric two-sample testing using the sliced Wasserstein (SW) distance. While prior theoretical and empirical work indicates that the SW distance offers a promising balance between strong statistical guarantees and computational efficiency, its theoretical foundations for hypothesis testing remain limited. We address this gap by proposing a permutation-based SW test and analyzing its performance. The test inherits finite-sample Type I error control from the permutation principle. Moreover, we establish non-asymptotic power bounds and show that the procedure achieves the minimax separation rate $n^{-1/2}$ over multinomial and bounded-support alternatives, matching the optimal guarantees of kernel-based tests while building on the geometric foundations of Wasserstein distances. Our analysis further quantifies the trade-off between the number of projections and statistical power. Finally, numerical experiments demonstrate that the test combines finite-sample validity with competitive power and scalability, and -- unlike kernel-based tests, which require careful kernel tuning -- it performs consistently well across all scenarios we consider.
format Preprint
id arxiv_https___arxiv_org_abs_2510_27498
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Minimax-Optimal Two-Sample Test with Sliced Wasserstein
Tran, Binh Thuan
Schreuder, Nicolas
Machine Learning
Statistics Theory
Methodology
We study the problem of nonparametric two-sample testing using the sliced Wasserstein (SW) distance. While prior theoretical and empirical work indicates that the SW distance offers a promising balance between strong statistical guarantees and computational efficiency, its theoretical foundations for hypothesis testing remain limited. We address this gap by proposing a permutation-based SW test and analyzing its performance. The test inherits finite-sample Type I error control from the permutation principle. Moreover, we establish non-asymptotic power bounds and show that the procedure achieves the minimax separation rate $n^{-1/2}$ over multinomial and bounded-support alternatives, matching the optimal guarantees of kernel-based tests while building on the geometric foundations of Wasserstein distances. Our analysis further quantifies the trade-off between the number of projections and statistical power. Finally, numerical experiments demonstrate that the test combines finite-sample validity with competitive power and scalability, and -- unlike kernel-based tests, which require careful kernel tuning -- it performs consistently well across all scenarios we consider.
title Minimax-Optimal Two-Sample Test with Sliced Wasserstein
topic Machine Learning
Statistics Theory
Methodology
url https://arxiv.org/abs/2510.27498