Two-sample statistics based on anisotropic kernels

DOI10.1093/IMAIAI/IAZ018MaRDI QIDQ5006535zbMATH OpenOpenAlexWikidataFDO

Authors Xiuyuan Cheng, Alexander Cloninger, Ronald R. Coifman

Publication date 16 August 2021

Published in Information and Inference: A Journal of the IMA (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1709.05006, https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7478116

zbMATH Keywords

anisotropic kernelmaximum mean discrepancytwo-sample statistics

Mathematics Subject Classification ID

Hilbert spaces with reproducing kernels (= (proper) functional Hilbert spaces, including de Branges-Rovnyak and other structured spaces) (46E22) Asymptotic properties of nonparametric inference (62G20) Hypothesis testing in multivariate analysis (62H15) Image analysis in multivariate analysis (62H35)

Abstract: The paper introduces a new kernel-based Maximum Mean Discrepancy (MMD) statistic for measuring the distance between two distributions given finitely-many multivariate samples. When the distributions are locally low-dimensional, the proposed test can be made more powerful to distinguish certain alternatives by incorporating local covariance matrices and constructing an anisotropic kernel. The kernel matrix is asymmetric; it computes the affinity between $n$ data points and a set of $n_R$ reference points, where $n_R$ can be drastically smaller than $n$. While the proposed statistic can be viewed as a special class of Reproducing Kernel Hilbert Space MMD, the consistency of the test is proved, under mild assumptions of the kernel, as long as $|p-q|sqrtnoinfty$, and a finite-sample lower bound of the testing power is obtained. Applications to flow cytometry and diffusion MRI datasets are demonstrated, which motivate the proposed approach to compare distributions.

Recommendations

• A kernel two-sample test
• EuMMD: efficiently computing the MMD two-sample test statistic for univariate data
• Two-sample test statistics for measuring discrepancies between two multivariate probability density functions using kernel-based density estimates
• Interpoint distance based two sample tests in high dimension
• Equivalence of distance-based and RKHS-based statistics in hypothesis testing

Cited in

(13)

• Two-sample test statistics for measuring discrepancies between two multivariate probability density functions using kernel-based density estimates
• Applied harmonic analysis and data processing. Abstracts from the workshop held March 25--31, 2018
• Supervised learning of sheared distributions using linearized optimal transport
• Local significant differences from nonparametric two-sample tests
• Characteristic and universal tensor product kernels
• A kernel two-sample test
• EuMMD: efficiently computing the MMD two-sample test statistic for univariate data
• Kernel two-sample tests for manifold data
• A new maximum mean discrepancy based two-sample test for equal distributions in separable metric spaces
• A distance based two-sample test of means difference for multivariate datasets
• Methods for quantifying dataset similarity: a review, taxonomy and comparison
• Asymptotic normality of a generalized maximum mean discrepancy estimator
• Equivalence of distance-based and RKHS-based statistics in hypothesis testing

This page was built for publication: Two-sample statistics based on anisotropic kernels

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5006535)