Robust modifications of U-statistics and applications to covariance estimation problems

DOI10.3150/19-BEJ1149MaRDI QIDQ2278677zbMATH OpenOpenAlexFDO

Authors Stanislav Minsker, Xiaohan Wei

Publication date 5 December 2019

Published in Bernoulli (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1801.05565, https://projecteuclid.org/euclid.bj/1574758843

zbMATH Keywords

covariance estimationheavy tailsrobust estimatorsU-statistics

Mathematics Subject Classification ID

Order statistics; empirical distribution functions (62G30) Statistics of extreme values; tail inference (62G32) Nonparametric robustness (62G35) Estimation in multivariate analysis (62H12) Analysis of variance and covariance (ANOVA) (62J10)

Abstract: Let $Y$ be a $d$-dimensional random vector with unknown mean $mu$ and covariance matrix $Sigma$. This paper is motivated by the problem of designing an estimator of $Sigma$ that admits tight deviation bounds in the operator norm under minimal assumptions on the underlying distribution, such as existence of only 4th moments of the coordinates of $Y$. To address this problem, we propose robust modifications of the operator-valued U-statistics, obtain non-asymptotic guarantees for their performance, and demonstrate the implications of these results to the covariance estimation problem under various structural assumptions.

Review text

Consider a random vector \(Y\) with unknown mean and unknown covariance matrix \(\Sigma.\) Independent copies of \(Y\) are observed. In the paper under review, the problem of designing an estimator of \(\Sigma\) is studied that admits exponential deviation bounds in the operator norm under mild assumptions, such as the existence of only the 4th moment of \(Y\). For this purpose, robust modifications of the operator-valued U-statistics are proposed and non-asymptotic results are obtained for their performance. Then implications for the covariance estimation problem and its versions are outlined.

Recommendations

• Robust covariance estimation under \(L_4\)-\(L_2\) norm equivalence
• Minimax estimation of large covariance matrices under _1-norm
• Structured robust covariance estimation
• Sub-Gaussian estimators of the mean of a random matrix with heavy-tailed entries
• Sub-Gaussian estimators of the mean of a random vector

Cites work

• A Class of Statistics with Asymptotically Normal Distribution
• A distribution-free M-estimator of multivariate scatter
• Adaptive estimation of the copula correlation matrix for semiparametric elliptical copulas
• Asymptotically Minimax Adaptive Estimation. I: Upper Bounds. Optimally Adaptive Estimates
• Challenging the empirical mean and empirical variance: a deviation study
• Concentration inequalities and moment bounds for sample covariance operators
• Convex optimization theory.
• Fast probabilistic algorithms for Hamiltonian circuits and matchings
• High-breakdown robust multivariate methods
• High-dimensional covariance matrix estimation with missing observations
• scientific article; zbMATH DE number 3790208 (Why is no real title available?)
• scientific article; zbMATH DE number 1254560 (Why is no real title available?)
• scientific article; zbMATH DE number 1256715 (Why is no real title available?)
• scientific article; zbMATH DE number 1552476 (Why is no real title available?)
• Limit theorems for \(U\)-processes
• Marčenko-Pastur law for Tyler's M-estimator
• On some extensions of Bernstein's inequality for self-adjoint operators
• Operator Lipschitz functions
• Partial estimation of covariance matrices
• Principle component analysis: robust versions
• Probability Inequalities for Sums of Bounded Random Variables
• Robust estimation of U-statistics
• Robust Estimation of a Location Parameter
• Robust m-estimators of multivariate location and scatter
• Robust principal component analysis?
• Statistical analysis of latent generalized correlation matrix estimation in transelliptical distribution
• Sub-Gaussian estimators of the mean of a random matrix with heavy-tailed entries
• Sub-Gaussian mean estimators
• The masked sample covariance estimator: an analysis using matrix concentration inequalities
• Trace inequalities and quantum entropy: an introductory course
• User-friendly tail bounds for sums of random matrices

Cited in

(15)

• Statistical inference for structured high-dimensional models. Abstracts from the workshop held March 11--17, 2018
• Approximating \(L_p\) unit balls via random sampling
• Robust and efficient mean estimation: an approach based on the properties of self-normalized sums
• New challenges in covariance estimation: multiple structures and coarse quantization
• Robust covariance estimation for distributed principal component analysis
• Robust covariance estimation under \(L_4\)-\(L_2\) norm equivalence
• Moment inequalities for matrix-valued U-statistics of order 2
• User-friendly covariance estimation for heavy-tailed distributions
• Robust estimation of U-statistics
• scientific article; zbMATH DE number 7370530 (Why is no real title available?)
• Understanding Implicit Regularization in Over-Parameterized Single Index Model
• Mean estimation in high dimension
• Gaussian differentially private robust mean estimation and inference
• Robust estimation of covariance matrices: adversarial contamination and beyond
• Improved covariance estimation: optimal robustness and sub-Gaussian guarantees under heavy tails

This page was built for publication: Robust modifications of U-statistics and applications to covariance estimation problems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2278677)