Spectral analysis of the Gram matrix of mixture models

DOI10.1051/PS/2016007MaRDI QIDQ2954233zbMATH OpenOpenAlexFDO

Authors Florent Benaych-Georges, R. Couillet

Publication date 12 January 2017

Published in ESAIM: Probability and Statistics (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1510.03463

zbMATH Keywords

mixture modelsspectral clusteringrandom matricesextreme eigenvalue statistics

Mathematics Subject Classification ID

Classification and discrimination; cluster analysis (statistical aspects) (62H30) Random matrices (algebraic aspects) (15B52) Random matrices (probabilistic aspects) (60B20)

Abstract: This text is devoted to the asymptotic study of some spectral properties of the Gram matrix $W^{sfT}W$ built upon a collection $w_1,ldots,w_{n}inmathbb{R}^p$ of random vectors (the columns of $W$), as both the number $n$ of observations and the dimension $p$ of the observations tend to infinity and are of similar order of magnitude. The random vectors $w_1,ldots,w_n$ are independent observations, each of them belonging to one of $k$ classes $mathcal{C}_1,ldots,mathcal{C}_k$. The observations of each class $mathcal{C}_a$ ($1lealek$) are characterized by their distribution $mathcalN(0,p^{-1}{C}_a)$, where $C_1,ldots,C_k$ are some non negative definite $pimesp$ matrices. The cardinality $n_a$ of class $mathcal{C}_a$ and the dimension $p$ of the observations are such that $frac{n}_{a}n$ ($1lealek$) and $fracpn$ stay bounded away from $0$ and $+infty$. We provide deterministic equivalents to the empirical spectral distribution of $W^{sfT}W$ and to the matrix entries of its resolvent (as well as of the resolvent of $W{W}^{sfT}$). These deterministic equivalents are defined thanks to the solutions of a fixed-point system. Besides, we prove that $W^{sfT}W$ has asymptotically no eigenvalues outside the bulk of its spectrum, defined thanks to these deterministic equivalents. These results are directly used in our companion paper "Kernel spectral clustering of large dimensional data", which is devoted to the analysis of the spectral clustering algorithm in large dimensions. They also find applications in various other fields such as wireless communications where functionals of the aforementioned resolvents allow one to assess the communication performance across multi-user multi-antenna channels.

Recommendations

• Kernel spectral clustering of large dimensional data
• scientific article; zbMATH DE number 5020283
• Covariance discriminative power of kernel clustering methods
• Spectral convergence for a general class of random matrices
• Spectral distribution of large generalized random kernel matrices

Cites work

• scientific article; zbMATH DE number 194139 (Why is no real title available?)
• scientific article; zbMATH DE number 6125590 (Why is no real title available?)
• A Central Limit Theorem for the SINR at the LMMSE Estimator Output for Large-Dimensional Signals
• A Deterministic Equivalent for the Analysis of Correlated MIMO Multiple Access Channels
• A concentration inequality and a local law for the sum of two random matrices
• Additive/multiplicative free subordination property and limiting eigenvectors of spiked additive deformations of Wigner matrices and spiked sample covariance matrices
• Almost sure localization of the eigenvalues in a Gaussian information plus noise model. Application to the spiked models.
• An introduction to statistical learning. With applications in R
• Analysis of the limiting spectral distribution of large dimensional random matrices
• Analysis of the limiting spectral measure of large random matrices of the separable covariance type
• DISTRIBUTION OF EIGENVALUES FOR SOME SETS OF RANDOM MATRICES
• Kernel spectral clustering of large dimensional data
• No eigenvalues outside the support of the limiting spectral distribution of large-dimensional sample covariance matrices
• On the distribution of the largest eigenvalue in principal components analysis
• Phase transition of the largest eigenvalue for nonnull complex sample covariance matrices
• Semicircle law on short scales and delocalization of eigenvectors for Wigner random matrices
• Spectral algorithms
• The eigenvalues and eigenvectors of finite, low rank perturbations of large random matrices
• The outliers among the singular values of large rectangular random matrices with additive fixed rank deformation
• The singular values and vectors of low rank perturbations of large rectangular random matrices

Cited in

(7)

• A method of spectral mixture analysis based on the Gaussian Markov random field model
• MIXANDMIX: numerical techniques for the computation of empirical spectral distributions of population mixtures
• The Spectral Method for General Mixture Models
• scientific article; zbMATH DE number 7370611 (Why is no real title available?)
• Covariance discriminative power of kernel clustering methods
• Optimal cleaning for singular values of cross-covariance matrices
• Eigenvalue distribution of some nonlinear models of random matrices

Describes a project that uses

Uses Software

• ISLR
• ElemStatLearn

This page was built for publication: Spectral analysis of the Gram matrix of mixture models

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2954233)