Randomized core reduction for discrete ill-posed problem (Q2309262): Difference between revisions

This paper focuses on solving of a linear approximation problem \(Ax\approx b\), where \(A\) is a matrix representing some linear mapping and \(b\) a right-hand side vector that is not in the range of the mapping. This incompatibility is due to errors that are contained in both \(A\) as well as \(b\) and therefore an approximate solution \(x\) is sought by using the total least squares (TLS) minimization. The TLS minimization can be done via the so-called core problem (CP) reduction. The main goal of this paper is an extraction of the CP from given data \(\{A,b\}\). Moreover, the authors consider the problem \(Ax\approx b\) to be a large-scale discrete ill-posed problem, which disqualifies the usage of standard methods. Therefore, the authors propose a randomized algorithm for CP extraction and give a basic error analysis for their approach. The efficiency of the randomized approach is illustrated by several numerical experiments.