Remote Access Mathematics of Computation
Green Open Access

Mathematics of Computation

ISSN 1088-6842(online) ISSN 0025-5718(print)

 
 

 

Backward error and conditioning of Fiedler companion linearizations


Author: Fernando De Terán
Journal: Math. Comp. 89 (2020), 1259-1300
MSC (2010): Primary 15A18, 65F15, 65F35
DOI: https://doi.org/10.1090/mcom/3480
Published electronically: October 17, 2019
Full-text PDF
View in AMS MathViewer New

Abstract | References | Similar Articles | Additional Information

Abstract: The standard way to solve polynomial eigenvalue problems is through linearizations. The family of Fiedler linearizations, which includes the classical Frobenius companion forms, presents many interesting properties from both the theoretical and the applied point of view. These properties make the Fiedler pencils a very attractive family of linearizations to be used in the solution of polynomial eigenvalue problems. However, their numerical features for general matrix polynomials had not yet been fully investigated. In this paper, we analyze the backward error of eigenpairs and the condition number of eigenvalues of Fiedler linearizations in the solution of polynomial eigenvalue problems. We get bounds for: (a) the ratio between the backward error of an eigenpair of the matrix polynomial and the backward error of the corresponding (computed) eigenpair of the linearization, and (b) the ratio between the condition number of an eigenvalue in the linearization and the condition number of the same eigenvalue in the matrix polynomial. A key quantity in these bounds is $ \rho $, the ratio between the maximum norm of the coefficients of the polynomial and the minimum norm of the leading and trailing coefficient. If the matrix polynomial is well scaled (i. e., all its coefficients have a similar norm, which implies $ \rho \approx 1$), then solving the Polynomial Eigenvalue Problem with any Fiedler linearization will give a good performance from the point of view of backward error and conditioning. In the more general case of badly scaled matrix polynomials, dividing the coefficients of the polynomial by the maximum norm of its coefficients allows us to get better bounds. In particular, after this scaling, the ratio between the eigenvalue condition number in any two Fiedler linearizations is bounded by a quantity that depends only on the size and the degree of the polynomial. We also analyze the effect of parameter scaling in these linearizations, which improves significantly the backward error and conditioning in some cases where $ \rho $ is large. Several numerical experiments are provided to support our theoretical results.


References [Enhancements On Off] (What's this?)


Similar Articles

Retrieve articles in Mathematics of Computation with MSC (2010): 15A18, 65F15, 65F35

Retrieve articles in all journals with MSC (2010): 15A18, 65F15, 65F35


Additional Information

Fernando De Terán
Affiliation: Departamento de Matemáticas, Universidad Carlos III de Madrid, Avda. Universidad 30, 28911 Leganés, Spain
Email: fteran@math.uc3m.es

DOI: https://doi.org/10.1090/mcom/3480
Keywords: Matrix polynomial, matrix pencil, eigenvalue, eigenvector, Polynomial Eigenvalue Problem, companion linearization, Fiedler pencil, conditioning, backward error, scaling
Received by editor(s): February 22, 2019
Received by editor(s) in revised form: June 11, 2019
Published electronically: October 17, 2019
Additional Notes: This work was partially supported by the Ministerio de Ciencia e Innovación of Spain through grant MTM-2009-09281, and by the Ministerio de Economía y Competitividad of Spain through grants MTM-2012-32542, MTM2015-68805-REDT, and MTM2015-65798-P
Article copyright: © Copyright 2019 American Mathematical Society