Ignace Loris

  1. On the applicability of compressed sensing to ill-conditioned and noisy systems.

    Authors: Ignace Loris, Caroline Verhoeven
    Subjects: Numerical Analysis
    Abstract

    The influence of noise and the effects of ill-conditioning of the measurement
    matrix on the effectiveness of $\ell_1$-based recovery of sparse signals in
    high dimensional spaces is investigated. Even for moderately ill-conditioned
    systems it is found that such a method performs much worse than for random
    matrix based problems that are often presented. On the other hand, when
    considering different noise levels for a fixed condition number, the relative
    accuracy remains constant.

RSS-материал