Spectral analysis of the Gram matrix of mixture models

Abstract : This text is devoted to the asymptotic study of some spectral properties of the Gram matrix WTW built upon a collection w1,...,wn ∈ Rp 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 w1,...,wn are independent observations, each of them belonging to one of k classes . The observations of each class (1 ≤ a ≤ k) are characterized by their distribution , where C1,...,Ck are some non negative definite p × p matrices. The cardinality na of class and the dimension p of the observations are such that na/n (1 ≤ a ≤ k) and p/n stay bounded away from 0 and + ∞. We provide deterministic equivalents to the empirical spectral distribution of WTW and to the matrix entries of its resolvent (as well as of the resolvent of WWT). These deterministic equivalents are defined thanks to the solutions of a fixed-point system. Besides, we prove that WTW 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 [R. Couillet and F. Benaych-Georges, Electron. J. Stat. 10 (2016) 1393–1454.], 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.
Type de document :
Article dans une revue
ESAIM: Probability and Statistics, EDP Sciences, 2016, 20, pp.217 - 237. 〈10.1051/ps/2016007〉
Liste complète des métadonnées

https://hal.archives-ouvertes.fr/hal-01215342
Contributeur : Florent Benaych-Georges <>
Soumis le : mercredi 14 octobre 2015 - 06:05:34
Dernière modification le : jeudi 31 mai 2018 - 09:12:02

Lien texte intégral

Identifiants

Citation

Florent Benaych-Georges, Romain Couillet. Spectral analysis of the Gram matrix of mixture models. ESAIM: Probability and Statistics, EDP Sciences, 2016, 20, pp.217 - 237. 〈10.1051/ps/2016007〉. 〈hal-01215342〉

Partager

Métriques

Consultations de la notice

260