Restricted isometry property for random matrices with heavy tailed columns

C R Acad Sci Paris Ser I ?? Contents lists available at ScienceDirect C R Acad Sci Paris Ser I www sciencedirect com Functional analysis Probability theory Restricted isometry property for random matrices with heavy-tailed columns Propriété d ? isométrie restreinte de matrices aléatoires dont les colonnes sont à queues lourdes Olivier Guédon a Alexander E Litvak b Alain Pajor a Nicole Tomczak-Jaegermann b a Université Paris-Est Laboratoire d ? analyse et de mathématiques appliquées UMR UPEMLV Marne-la-Vallée France b Department of Mathematical and Statistical Sciences University of Alberta Edmonton Alberta T G G Canada article info Article history Received January Accepted March Available online March Presented by Gilles Pisier abstract Let A be a matrix whose columns X XN are independent random vectors in Rn Assume that p-th moments of Xi a a ?? Sn ?? i N are uniformly bounded For p we prove that with high probability A has the Restri ??cted Isometry Property RIP provided that Euclidean norms Xi are concentrated around n and that the covariance matrix is well approximated by the empirical covariance matrix provided that maxi Xi C nN We also provide estimates for RIP when E Xi a for t exp t with ?? ? Published by Elsevier Masson SAS on behalf of Académie des sciences résumé Soit A une matrice dont les colonnes X XN sont des vecteurs indépendants de Rn On suppose que les moments d ? ordre p des Xi a a ?? Sn ?? i N sont uniformément bornés pour ??p On démontre que si les normes euclidiennes des Xi se concentrent autour de n la matrice A véri ?e une propriété d ? isométrie restreinte avec grande probabilité et que si maxi Xi C nN la matrice de covariance empirique est une bonne approximation de la matrice de covariance On démontre aussi une propriété d ? isométrie restreinte quand E Xi a pour tout a ?? Sn ?? i N avec t exp t et ?? ? Published by Elsevier Masson SAS on behalf of Académie des sciences E-mail addresses olivier guedon u-pem fr O Guédon aelitvak gmail com A E Litvak alain pajor u-pem fr A Pajor nicole tomczak ualberta ca N Tomczak-Jaegermann This author holds the Canada Research Chair in Geometric Analysis http dx doi org j crma - X ? Published by Elsevier Masson SAS on behalf of Académie des sciences C O Guédon et al C R Acad Sci Paris Ser I ?? Introduction Our two main results go in two parallel directions the Restricted Isometry Property abbreviated as RIP and a question of Kannan ??Lovász ??Simonovits about an approximation of a covariance matrix by empirical covariance matrices referred to below as KLS problem In this note X XN denote independent random vectors in Rn satisfying for some function ?? i N ??a ?? Sn ?? E Xi a We will focus on two choices of the function t t p with p or t exp t with ?? The n ? N

  • 31
  • 0
  • 0
Afficher les détails des licences
Licence et utilisation
Gratuit pour un usage personnel Aucune attribution requise
Partager