Computing eigenvalues of quasi-generalized Vandermonde matrices to high relative accuracy

Abstract

In this paper, the eigenvalue problem for the class of quasi-generalized Vandermonde (q-gV) matrices is considered. In order to parameterize q-gV matrices, the explicit expressions of minors of such matrices are presented. We develop an algorithm to accurately compute the parameterization for q-gV matrices. Relying on the accurate parameterization, all the eigenvalues of q-gV matrices are computed to high relative accuracy. Error analysis and numerical experiments are provided to confirm the high relative accuracy.