Current browse context:
math.ST
Change to browse by:
References & Citations
Mathematics > Statistics Theory
Title: Restricted Isometry Constants for Gaussian and Rademacher matrices
(Submitted on 5 Apr 2016 (v1), revised 1 Mar 2017 (this version, v2), latest version 14 Nov 2018 (v4))
Abstract: Restricted Isometry Constants (RICs) are a pivotal notion in Compressed Sensing as these constants finely assess how a linear operator is conditioned on the set of sparse vectors and hence how it performs in stable and robust sparse regression (SRSR). While it is an open problem to construct deterministic matrices with apposite RICs, one can prove that such matrices exist using random matrices models. In this paper, we show upper bounds on RICs for Gaussian and Rademacher matrices using state-of-the-art small deviation estimates on their extreme eigenvalues. This allows us to derive a lower bound on the probability of getting SRSR. One of the benefits of this approach is to introduce a simple tool from Random Matrix Theory to derive upper bounds on RICs and phase transition on SRSR from small deviations on the extreme eigenvalues.
Submission history
From: Yohann De Castro [view email][v1] Tue, 5 Apr 2016 08:37:20 GMT (615kb)
[v2] Wed, 1 Mar 2017 10:36:44 GMT (188kb,D)
[v3] Mon, 4 Dec 2017 10:47:46 GMT (163kb,D)
[v4] Wed, 14 Nov 2018 12:34:01 GMT (165kb,D)
Link back to: arXiv, form interface, contact.