Restricted isometry property for random matrices with heavy tailed columns
نویسندگان
چکیده
Let A be a matrix whose columns X1, . . . , XN are independent random vectors in R. Assume that p-th moments of 〈Xi, a〉, a ∈ Sn−1, i ≤ N , are uniformly bounded. For p > 4 we prove that with high probability A has the Restricted 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 Eφ (|〈Xi, a〉|) ≤ 1 for φ(t) = (1/2) exp(t), with α ∈ (0, 2]. Soit A une matrice dont les colonnes X1, . . . , XN sont des vecteurs indépendants de R. On suppose que les moments d’ordre p des 〈Xi, a〉, a ∈ Sn−1, 1 ≤ i ≤ N sont uniformément bornés pour un p > 4. On démontre que si les normes euclidiennes des |Xi| se concentrent autour de √ n , la matrice A vérifie 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 Eφ (|〈Xi, a〉|) ≤ 1 pour tout a ∈ Sn−1, 1 ≤ i ≤ N avec φ(t) = (1/2) exp(t) et α ∈ (0, 2].
منابع مشابه
Restricted isometry property of matrices with independent columns and neighborly polytopes by random sampling
This paper considers compressed sensing matrices and neighborliness of a centrally symmetric convex polytope generated by vectors ±X1, . . . ,±XN ∈ Rn, (N ≥ n). We introduce a class of random sampling matrices and show that they satisfy a restricted isometry property (RIP) with overwhelming probability. In particular, we prove that matrices with i.i.d. centered and variance 1 entries that satis...
متن کاملImproved Bounds on Restricted Isometry Constants for Gaussian Matrices
The Restricted Isometry Constants (RIC) of a matrix A measures how close to an isometry is the action of A on vectors with few nonzero entries, measured in the `2 norm. Specifically, the upper and lower RIC of a matrix A of size n ×N is the maximum and the minimum deviation from unity (one) of the largest and smallest, respectively, square of singular values of all `N k ́ matrices formed by taki...
متن کاملRobustness Properties of Dimensionality Reduction with Gaussian Random Matrices
In this paper we study the robustness properties of dimensionality reduction with Gaussian random matrices having arbitrarily erased rows. We first study the robustness property against erasure for the almost norm preservation property of Gaussian random matrices by obtaining the optimal estimate of the erasure ratio for a small given norm distortion rate. As a consequence, we establish the rob...
متن کاملImproved Bounds on Restricted Isometry Constants
The restricted isometry constant (RIC) of a matrix A measures how close to an isometry is the action of A on vectors with few nonzero entries, measured in the 2 norm. Specifically, the upper and lower RICs of a matrix A of size n×N are the maximum and the minimum deviation from unity (one) of the largest and smallest, respectively, square of singular values of all (N k ) matrices formed by taki...
متن کاملOpen Problem: Restricted Eigenvalue Condition for Heavy Tailed Designs
The restricted eigenvalue (RE) condition characterizes the sample complexity of accurate recovery in the context of high-dimensional estimators such as Lasso and Dantzig selector (Bickel et al., 2009). Recent work has shown that random design matrices drawn from any thin-tailed (subGaussian) distributions satisfy the RE condition with high probability, when the number of samples scale as the sq...
متن کامل