Robust covariance estimation under L₄-L₂ norm equivalence

DOI10.1214/19-AOS1862zbMATH Open1451.62084arXiv1809.10462OpenAlexW3043175749MaRDI QIDQ2196239FDOQ2196239

Shahar Mendelson, Nikita Zhivotovskiy

Publication date: 28 August 2020

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

Abstract: Let $X$ be a centered random vector taking values in $mathbb{R}^d$ and let $Sigma=mathbbE(XotimesX)$ be its covariance matrix. We show that if $X$ satisfies an $L_4-L_2$ norm equivalence, there is a covariance estimator $hatSigma$ that exhibits the optimal performance one would expect had $X$ been a gaussian vector. The procedure also improves the current state-of-the-art regarding high probability bounds in the subgaussian case (sharp results were only known in expectation or with constant probability). In both scenarios the new bound does not depend explicitly on the dimension $d$, but rather on the effective rank of the covariance matrix $Sigma$.

Full work available at URL: https://arxiv.org/abs/1809.10462

Recommendations

• Robust modifications of U-statistics and applications to covariance estimation problems
• Estimating covariance and precision matrices along subspaces
• Sub-Gaussian estimators of the mean of a random matrix with heavy-tailed entries
• Structured robust covariance estimation
• Robust covariance and scatter matrix estimation under Huber's contamination model

zbMATH Keywords

covariance estimationrobust estimationmedian of means

Mathematics Subject Classification ID

Nonparametric robustness (62G35) Nonparametric tolerance and confidence regions (62G15) Analysis of variance and covariance (ANOVA) (62J10)

Cites Work

• Multivariate T-Distributions and Their Applications
• Challenging the empirical mean and empirical variance: a deviation study
• High-dimensional covariance matrix estimation with missing observations
• Title not available (Why is that?)
• An Introduction to Matrix Concentration Inequalities
• Upper and Lower Bounds for Stochastic Processes
• Robust modifications of U-statistics and applications to covariance estimation problems
• Reconstruction and subgaussian operators in asymptotic geometric analysis
• Upper bounds on product and multiplier empirical processes
• Empirical processes with a bounded \(\psi_1\) diameter
• On the infimum convolution inequality
• Concentration inequalities and moment bounds for sample covariance operators
• On some extensions of Bernstein's inequality for self-adjoint operators
• Sub-Gaussian estimators of the mean of a random matrix with heavy-tailed entries
• Facial structures of schattenp-Norms
• Near-optimal mean estimators with respect to general norms

Cited In (27)

• New challenges in covariance estimation: multiple structures and coarse quantization
• Covariance estimation under one-bit quantization
• Finite-sample analysis of \(M\)-estimators using self-concordance
• Sub-Gaussian estimators of the mean of a random vector
• Suboptimality of constrained least squares and improvements via non-linear predictors
• Uniform Hanson-Wright type concentration inequalities for unbounded entries via the entropy method
• Robustifying Markowitz
• A new robust covariance matrix estimation for high-dimensional microbiome data
• The geometric median and applications to robust mean estimation
• Gaussian differentially private robust mean estimation and inference
• Title not available (Why is that?)
• Robust subgaussian estimation with VC-dimension
• Robust estimation of covariance matrices: adversarial contamination and beyond
• Covariance estimation under missing observations and \(L_4 - L_2\) moment equivalence
• Stable recovery and the coordinate small-ball behaviour of random vectors
• Distribution-free robust linear regression
• Quantitative limit theorems and bootstrap approximations for empirical spectral projectors
• Tyler's and Maronna's M-estimators: non-asymptotic concentration results
• Robust \(k\)-means clustering for distributions with two moments
• Dimension-free bounds for sums of independent matrices and simple tensors via the variational principle
• Approximating \(L_p\) unit balls via random sampling
• Robust covariance estimation for distributed principal component analysis
• Nearly optimal robust mean estimation via empirical characteristic function
• Rank-based correlation matrix estimation for high dimensional microbiome data
• Mean estimation and regression under heavy-tailed distributions: A survey
• Improved covariance estimation: optimal robustness and sub-Gaussian guarantees under heavy tails
• User-friendly covariance estimation for heavy-tailed distributions

This page was built for publication: Robust covariance estimation under \(L_4\)-\(L_2\) norm equivalence

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2196239)