Quantile-based Random Sparse Kaczmarz for Corrupted and Noisy Linear Systems

arXiv2206.07356MaRDI QIDQ6402147FDOQ6402147

Lu Zhang, Hui Zhang, Hongxia Wang

Publication date: 15 June 2022

Abstract: It is common for us to meet with large-scale corrupted and noisy linear inverse problems in practical applications. In this study, we propose a quantile-based randomized sparse Kaczmarz method, based on recently proposed nonsmooth Bregman projection and quantile techniques, for obtaining sparse solutions to corrupted and noisy linear inverse problems. Moreover, we incorporate the heavy ball momentum technique into the proposed method for possible speedup. Theoretically, we show that the quantile-based randomized sparse Kaczmarz method converges linearly in expectation with respect to the Bregman distance. Numerically, we demonstrate the superiority of our proposed algorithms through extensive numerical experiments.

This page was built for publication: Quantile-based Random Sparse Kaczmarz for Corrupted and Noisy Linear Systems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6402147)