On mixed and componentwise condition numbers for indefinite least squares problem

Abstract

In this paper, we mainly consider the mixed and componentwise condition numbers for the indefinite least squares problem without constraints or with least squares constraints. The explicit expressions and the easily computable upper bounds for these condition numbers are presented. Two numerical examples are given to test these results and illustrate their superiority compared with the corresponding normwise condition numbers. Furthermore, we also provide the mixed and componentwise condition numbers for the Moore–Penrose inverse in indefinite inner product spaces.