Equivalence of distance-based and RKHS-based statistics in hypothesis testing

DOI10.1214/13-AOS1140zbMATH Open1281.62117arXiv1207.6076MaRDI QIDQ85652FDOQ85652

Bharath K. Sriperumbudur, Kenji Fukumizu, Kenji Fukumizu, Dino Sejdinovic, Dino Sejdinovic, Arthur Gretton, Arthur Gretton, Bharath Sriperumbudur

Publication date: 1 October 2013

Published in: The Annals of Statistics (Search for Journal in Brave)

Abstract: We provide a unifying framework linking two classes of statistics used in two-sample and independence testing: on the one hand, the energy distances and distance covariances from the statistics literature; on the other, maximum mean discrepancies (MMD), that is, distances between embeddings of distributions to reproducing kernel Hilbert spaces (RKHS), as established in machine learning. In the case where the energy distance is computed with a semimetric of negative type, a positive definite kernel, termed distance kernel, may be defined such that the MMD corresponds exactly to the energy distance. Conversely, for any positive definite kernel, we can interpret the MMD as energy distance with respect to some negative-type semimetric. This equivalence readily extends to distance covariance using kernels on the product space. We determine the class of probability distributions for which the test statistics are consistent against all alternatives. Finally, we investigate the performance of the family of distance kernels in two-sample and independence tests: we show in particular that the energy distance most commonly employed in statistics is just one member of a parametric family of kernels, and that other choices from this family can yield more powerful tests.

Full work available at URL: https://arxiv.org/abs/1207.6076

Recommendations

• The exact equivalence of distance and kernel methods in hypothesis testing
• A kernel two-sample test
• Characteristic and universal tensor product kernels
• Two-sample statistics based on anisotropic kernels
• A Hilbert Space Embedding for Distributions

zbMATH Keywords

distance covarianceindependence testingreproducing kernel Hilbert spacestwo-sample testing

Mathematics Subject Classification ID

Nonparametric hypothesis testing (62G10) Learning and adaptive systems in artificial intelligence (68T05) Applications of functional analysis in probability theory and statistics (46N30) Computational learning theory (68Q32)

Cites Work

• Measuring and testing dependence by correlation of distances
• Distance covariance in metric spaces
• A test for the two-sample problem based on empirical characteristic functions
• On the bootstrap of \(U\) and \(V\) statistics
• Two-sample test statistics for measuring discrepancies between two multivariate probability density functions using kernel-based density estimates
• On a new multivariate two-sample test.
• On the empirical estimation of integral probability metrics
• Consistent nonparametric tests of independence
• Hilbert space embeddings and metrics on probability measures
• 10.1162/153244303768966085
• Support Vector Machines
• A Hilbert Space Embedding for Distributions
• Title not available (Why is that?)
• Title not available (Why is that?)
• Integral Probability Metrics and Their Generating Classes of Functions
• Algorithmic Learning Theory
• Universality, Characteristic Kernels and RKHS Embedding of Measures
• Title not available (Why is that?)
• Title not available (Why is that?)
• Discussion of: Brownian distance covariance
• A new test for multivariate normality
• Brownian distance covariance

Cited In (only showing first 100 items - show all)

• An empirical study of the maximal and total information coefficients and leading measures of dependence
• Performance analysis of greedy algorithms for minimising a maximum mean discrepancy
• New efficient algorithms for multiple change-point detection with reproducing kernels
• Distance multivariance: new dependence measures for random vectors
• Four simple axioms of dependence measures
• Large-scale kernel methods for independence testing
• Distance-based and RKHS-based dependence metrics in high dimension
• Robust multivariate nonparametric tests via projection averaging
• Equivalence of kernel machine regression and kernel distance covariance for multidimensional phenotype association studies
• A significance test of the RV coefficient in high dimensions
• A kernel-based measure for conditional mean dependence
• On distance covariance in metric and Hilbert spaces
• An Updated Literature Review of Distance Correlation and Its Applications to Time Series
• A consistent test of independence based on a sign covariance related to Kendall's tau
• A generalization of an integral arising in the theory of distance correlation
• Multivariate Rank-Based Distribution-Free Nonparametric Testing Using Measure Transportation
• General tests of independence based on empirical processes indexed by functions
• A general Monte Carlo method for multivariate goodness-of-fit testing applied to elliptical families
• A one-sample test for normality with kernel methods
• Probabilistic sensitivity measures as information value
• A Hilbert Space Embedding for Distributions
• Comparing a large number of multivariate distributions
• Global sensitivity analysis with dependence measures
• Statistical inference on random dot product graphs: a survey
• High-dimensional variable screening through kernel-based conditional mean dependence
• Title not available (Why is that?)
• A basic treatment of the distance covariance
• Partial martingale difference correlation
• Bayesian Quadrature, Energy Minimization, and Space-Filling Design
• dcortools
• FastMMD: Ensemble of Circular Discrepancy for Efficient Two-Sample Test
• The distance standard deviation
• Conditional mean and quantile dependence testing in high dimension
• The exact equivalence of distance and kernel methods in hypothesis testing
• Partial distance correlation with methods for dissimilarities
• The Hellinger Correlation
• BET on Independence
• A Kernel Multiple Change-point Algorithm via Model Selection
• Stable correlation and robust feature screening
• Distance Metrics for Measuring Joint Dependence with Application to Causal Inference
• Expected similarity estimation for large-scale batch and streaming anomaly detection
• Expected Conditional Characteristic Function-based Measures for Testing Independence
• A new framework for distance and kernel-based metrics in high dimensions
• Statistical dependence: beyond Pearson's \(\rho\)
• Geometrical Insights for Implicit Generative Modeling
• A fast algorithm for computing distance correlation
• The BLUE in continuous-time regression models with correlated errors
• A Semiparametric Kernel Independence Test With Application to Mutational Signatures
• Distance correlation coefficients for Lancaster distributions
• Optimal rates for independence testing via $U$-statistic permutation tests
• Nonparametric tests for independence: a review and comparative simulation study with an application to malnutrition data in India
• On some consistent tests of mutual independence among several random vectors of arbitrary dimensions
• Title not available (Why is that?)
• A review of goodness-of-fit tests for models involving functional data
• Testing independence and goodness-of-fit jointly for functional linear models
• Title not available (Why is that?)
• Applications of distance correlation to time series
• A Generic Sure Independence Screening Procedure
• From Distance Correlation to Multiscale Graph Correlation
• Distribution-Free Consistent Independence Tests via Center-Outward Ranks and Signs
• On universally consistent and fully distribution-free rank tests of vector independence
• A regression perspective on generalized distance covariance and the Hilbert-Schmidt independence criterion
• Multivariate ranks and quantiles using optimal transport: consistency, rates and nonparametric testing
• Ball divergence: nonparametric two sample test
• Title not available (Why is that?)
• General tests of conditional independence based on empirical processes indexed by functions
• Rényi 100, quantitative and qualitative (in)dependence
• Distances defined by zonoids and statistical tests
• EuMMD: efficiently computing the MMD two-sample test statistic for univariate data
• Independence Weights for Causal Inference with Continuous Treatments
• Sufficient variable screening with high-dimensional controls
• Characteristic and Universal Tensor Product Kernels
• Self-supervised Metric Learning in Multi-View Data: A Downstream Task Perspective
• Title not available (Why is that?)
• Distribution of Distances based Object Matching: Asymptotic Inference
• Model predictivity assessment: incremental test-set selection and accuracy evaluation
• Measuring and testing homogeneity of distributions by characteristic distance
• Testing semiparametric model-equivalence hypotheses based on the characteristic function
• A double-robust test for high-dimensional gene coexpression networks conditioning on clinical information
• Title not available (Why is that?)
• Generalization of the HSIC and distance covariance using PDI kernels
• Methods for Assessment of Memory Reactivation
• Radon-Sobolev variational auto-encoders
• Sufficient variable selection using independence measures for continuous response
• Title not available (Why is that?)
• An Omnibus Non-Parametric Test of Equality in Distribution for Unknown Functions
• Evaluating Forecasts for High-Impact Events Using Transformed Kernel Scores
• Robust sufficient dimension reduction via ball covariance
• Test of conditional independence in factor models via Hilbert-Schmidt independence criterion
• Determinantal consensus clustering
• Power Analysis of Projection-Pursuit Independence Tests
• Antithetic and Monte Carlo kernel estimators for partial rankings
• Testing homogeneity in high dimensional data through random projections
• A Wavelet-Based Independence Test for Functional Data With an Application to MEG Functional Connectivity
• A consistent version of distance covariance for right‐censored survival data and its application in hypothesis testing
• BLUE against OLSE in the location model: energy minimization and asymptotic considerations
• Ranking Features to Promote Diversity: An Approach Based on Sparse Distance Correlation
• Asymptotic Distribution-Free Independence Test for High-Dimension Data
• A kernel- and optimal transport- based test of independence between covariates and right-censored lifetimes
• Kernel partial correlation: a novel approach to capturing conditional independence in graphical models for noisy data

This page was built for publication: Equivalence of distance-based and RKHS-based statistics in hypothesis testing

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q85652)