On the application of Laguerre's method to the polynomial eigenvalue problem

TL;DR

This paper introduces a Laguerre iteration-based method for solving polynomial eigenvalue problems, providing initial estimates, stability guarantees, and efficient eigenvector and condition number computations, with applications to structured matrix polynomials.

Contribution

It presents a novel Laguerre method for polynomial eigenvalue problems, including stability analysis, eigenvector computation, and efficiency improvements for structured matrices.

Findings

Method achieves accurate eigenvalues and eigenvectors.

Structured matrices like Hessenberg and tridiagonal benefit from computational savings.

Numerical experiments confirm competitiveness and stability of the approach.

Abstract

The polynomial eigenvalue problem arises in many applications and has received a great deal of attention over the last decade. The use of root-finding methods to solve the polynomial eigenvalue problem dates back to the work of Kublanovskaya (1969, 1970) and has received a resurgence due to the work of Bini and Noferini (2013). In this paper, we present a method which uses Laguerre iteration for computing the eigenvalues of a matrix polynomial. An effective method based on the numerical range is presented for computing initial estimates to the eigenvalues of a matrix polynomial. A detailed explanation of the stopping criteria is given, and it is shown that under suitable conditions we can guarantee the backward stability of the eigenvalues computed by our method. Then, robust methods are provided for computing both the right and left eigenvectors and the condition number of each eigenpair. Applications for Hessenberg and tridiagonal matrix polynomials are given and we show that both structures benefit from substantial computational savings. Finally, we present several numerical experiments to verify the accuracy of our method and its competitiveness for solving the roots of a polynomial and the tridiagonal eigenvalue problem.