Sharp bounds on the rate of convergence of the empirical covariance matrix

DOI10.1016/J.CRMA.2010.12.014MaRDI QIDQ627752zbMATH OpenOpenAlexWikidataFDO

Authors Radosław Adamczak, Alexander E. Litvak, Alain Pajor, Nicole Tomczak-Jaegermann

Publication date 3 March 2011

Published in Comptes Rendus. Mathématique. Académie des Sciences, Paris (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1012.0294

Mathematics Subject Classification ID

Random matrices (probabilistic aspects) (60B20) Inequalities; stochastic orderings (60E15) Strong limit theorems (60F15) Convexity and finite-dimensional Banach spaces (including special norms, zonoids, etc.) (aspects of convex geometry) (52A21) Probabilistic methods in Banach space theory (46B09)

Abstract: Let $X_1,\mathrm{...},X_{N}in{R}^n$ be independent centered random vectors with log-concave distribution and with the identity as covariance matrix. We show that with overwhelming probability at least $1-3exp(-csqrtn)$ one has $su{p}_{xin{S}^{n-1}}Big|frac1/Nsu{m}_{i=1}^N(|<X_i,x>{|}^2-E|<X_i,x>{|}^2)Big|leqCsqrtfracn/N,$ where $C$ is an absolute positive constant. This result is valid in a more general framework when the linear forms $(<X_i,x>{)}_{ileqN,xin{S}^{n-1}}$ and the Euclidean norms $(|X_i|/sqrtn{)}_{ileqN}$ exhibit uniformly a sub-exponential decay. As a consequence, if $A$ denotes the random matrix with columns $(X_i)$, then with overwhelming probability, the extremal singular values $lambd{a}_{mmin}$ and $lambd{a}_{mmax}$ of $A{A}^{o}p$ satisfy the inequalities $1-Csqrtn/Nlelambd{a}_{mmin}/Nlefraclambd{a}_{mmax}/Nle1+Csqrtn/N$ which is a quantitative version of Bai-Yin theorem cite{BY} known for random matrices with i.i.d. entries.

Recommendations

• On the convergence of the extremal eigenvalues of empirical covariance matrices with dependence
• Quantitative estimates of the convergence of the empirical covariance matrix in log-concave ensembles
• Strong Law for the eigenvalues and eigenvectors of empirical covariance matrices
• On the interval of fluctuation of the singular values of random matrices
• The rate of convergence of spectra of sample covariance matrices

Cites work

• Concentration of mass on convex bodies
• Isoperimetric problems for convex bodies and a localization lemma
• Limit of the smallest eigenvalue of a large dimensional sample covariance matrix
• Quantitative estimates of the convergence of the empirical covariance matrix in log-concave ensembles
• Random points in the unit ball of \(\ell^{ n }_{ p }\)
• Random walks and anO*(n5) volume algorithm for convex bodies
• Sampling convex bodies: a random matrix approach

Cited in

(23)

• Covariance estimation for distributions with \({2+\varepsilon}\) moments
• Large-dimensional central limit theorem with fourth-moment error bounds on convex sets and balls
• Coverings of random ellipsoids, and invertibility of matrices with i.i.d. heavy-tailed entries
• Finite impulse response models: a non-asymptotic analysis of the least squares estimator
• The lower tail of random quadratic forms with applications to ordinary least squares
• On the interval of fluctuation of the singular values of random matrices
• Random polytopes generated by contoured distributions
• Quantitative Version of a Silverstein’s Result
• Rank-one multi-reference factor analysis
• On the convergence of the extremal eigenvalues of empirical covariance matrices with dependence
• On the uniform convergence of empirical norms and inner products, with application to causal inference
• Bootstrapping the operator norm in high dimensions: error estimation for covariance matrices and sketching
• On generic chaining and the smallest singular value of random matrices with heavy tails
• Geometry of log-concave ensembles of random matrices and approximate reconstruction
• Concentration phenomena in high dimensional geometry
• Tyler's and Maronna's M-estimators: non-asymptotic concentration results
• Estimating the covariance of random matrices
• The smallest singular value of random rectangular matrices with no moment assumptions on entries
• How close is the sample covariance matrix to the actual covariance matrix?
• Quantitative estimates of the convergence of the empirical covariance matrix in log-concave ensembles
• Covariance estimation under one-bit quantization
• Restricted isometry property for random matrices with heavy-tailed columns
• Almost sure convergence of the largest and smallest eigenvalues of high-dimensional sample correlation matrices

This page was built for publication: Sharp bounds on the rate of convergence of the empirical covariance matrix

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q627752)