Approximating the covariance ellipsoid

DOI10.1142/S0219199719500895MaRDI QIDQ5120344zbMATH OpenOpenAlexWikidataFDO

Authors Shahar Mendelson

Publication date 10 September 2020

Published in Communications in Contemporary Mathematics (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1804.05402

zbMATH Keywords

covariance estimationhigh-dimensional statistics

Mathematics Subject Classification ID

Applications of functional analysis in probability theory and statistics (46N30) Measures of association (correlation, canonical correlation, etc.) (62H20) Analysis of variance and covariance (ANOVA) (62J10) Neural nets and related approaches to inference from stochastic processes (62M45)

Abstract: We explore ways in which the covariance ellipsoid $calB=vinmathbb{R}^d:mathbbE<X,v{>}^{2}leq1$ of a centred random vector $X$ in $mathbb{R}^d$ can be approximated by a simple set. The data one is given for constructing the approximating set consists of $X_1,\mathrm{...},X_N$ that are independent and distributed as $X$. We present a general method that can be used to construct such approximations and implement it for two types of approximating sets. We first construct a (random) set $calK$ defined by a union of intersections of slabs $H_{z,alpha}=vinmathbb{R}^d:|<z,v>|leqalpha$ (and therefore $calK$ is actually the output of a neural network with two hidden layers). The slabs are generated using $X_1,\mathrm{...},X_N$, and under minimal assumptions on $X$ (e.g., $X$ can be heavy-tailed) it suffices that $N=c_{1}det{a}^{-4}log(2/eta)$ to ensure that $(1-eta)calKsubsetcalBsubset(1+eta)calK$. In some cases (e.g., if $X$ is rotation invariant and has marginals that are well behaved in some weak sense), a smaller sample size suffices: $N=c_{1}det{a}^{-2}log(2/eta)$. We then show that if the slabs are replaced by randomly generated ellipsoids defined using $X_1,\mathrm{...},X_N$, the same degree of approximation is true when $Ngeq{c}_{2}det{a}^{-2}log(2/eta)$. The construction we use is based on the small-ball method.

Recommendations

• Covariance estimation for distributions with \({2+\varepsilon}\) moments
• Learning Theory
• Sampling convex bodies: a random matrix approach
• Quantitative estimates of the convergence of the empirical covariance matrix in log-concave ensembles
• On the Minimum Volume Covering Ellipsoid of Ellipsoids

Cites work

• A better approximation for balls
• Approximation of Convex Bodies by Polytopes
• Approximation of convex bodies by polytopes
• Concentration inequalities. A nonasymptotic theory of independence
• Covariance estimation for distributions with \({2+\varepsilon}\) moments
• Geometric applications of Chernoff-type estimates and a ZigZag approximation for balls
• scientific article; zbMATH DE number 49190 (Why is no real title available?)
• scientific article; zbMATH DE number 741240 (Why is no real title available?)
• scientific article; zbMATH DE number 1950575 (Why is no real title available?)
• Learning without concentration
• Neural Network Learning
• Noise stability of functions with low influences: invariance and optimality
• On the singular values of random matrices
• Probability. Theory and examples.
• Quantitative estimates of the convergence of the empirical covariance matrix in log-concave ensembles
• Sample Covariance Matrices of Heavy-Tailed Distributions
• Sharper bounds for Gaussian and empirical processes
• Some limit theorems for empirical processes (with discussion)
• Weak convergence and empirical processes. With applications to statistics

Cited in

(4)

• Approximating \(L_p\) unit balls via random sampling
• Mean estimation in high dimension
• Stable recovery and the coordinate small-ball behaviour of random vectors
• Structure preservation via the Wasserstein distance

This page was built for publication: Approximating the covariance ellipsoid

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5120344)