On the complex Falk-Langemeyer method

DOI10.1007/S11075-019-00689-8MaRDI QIDQ2290913zbMATH OpenOpenAlexFDO

Authors Vjeran Hari

Publication date 29 January 2020

Published in Numerical Algorithms (Search for Journal in Brave)

Full work available at URL https://doi.org/10.1007/s11075-019-00689-8

zbMATH Keywords

complex Hermitian matricesconditioningdefinite matrix pairdiagonalization methodgeneralized eigenvalue problem

Mathematics Subject Classification ID

Eigenvalues, singular values, and eigenvectors (15A18) Diagonalization, Jordan forms (15A20) Numerical computation of eigenvalues and eigenvectors of matrices (65F15) Numerical computation of matrix norms, conditioning, scaling (65F35)

Review text

The author presents an algorithm for a high relative accuracy computation of the generalized eigenvalue problem \(A x = \lambda B x\), \(x \neq 0\), for the case that the matrices \(A\) and \(B\) build a pair of definite complex Hermitian matrices. A pair of such matrices is definite if there exist real numbers \(c_1\) and \(c_2\) such that \(c_1 A + c_2 B\) is positive definite. The algorithm is an extension of the method of \textit{S. Falk} and \textit{P. Langemeyer} [Elektron. Datenverarb. No. 7, 30--34 (1960; Zbl 0100.33104)] for the simultaneous diagonalization of a pair \(A, B\) of real symmetric positive definite matrices. The author shows by numerical tests that the eigenvalues and eigenvectors are computed with high relative accuracy provided that the condition numbers of \(D_A A D_A\) and \(D_B B D_B\) are small for some diagonal matrices \(D_A\) and \(D_B\). The algorithm works also very well when the starting matrices are ill-conditioned with respect to matrix inversion. Another advantage is the inherent parallelism. The algorithm is very fast if \(A\) and \(B\) are nearly diagonal. Due to its properties the method can be used also as a kernel algorithm for block Jacobi methods and in the computation of generalized singular value decompositions. The complex Falk-Langemeyer algorithm is derived in great detail. A sequence of congruent matrix pairs is generated to solve the problem. The transformation matrices define the so-called pivot pairs which are selected by a pivot strategy. A special algorithm for the computation of the corresponding transformation parameters is presented. Some ideas for the proofs of the global and of the quadratic asymptotic convergence of the method are provided. Another topic of the paper is the analysis of the influence of rounding errors and an extensive section on the choice of numerical tests to validate the theoretical results.

Recommendations

• Accuracy of one step of the Falk-Langemeyer method
• On the Quadratic Convergence of the Falk–Langemeyer Method
• scientific article; zbMATH DE number 815446
• An algorithm for the symmetric generalized eigenvalue problem
• An Efficient and Stable Algorithm for the Symmetric-Definite Generalized Eigenvalue Problem

Cites work

• A One-Sided Jacobi Algorithm for Computing the Singular Value Decomposition on a Vector Computer
• A Tangent Algorithm for Computing the Generalized Singular Value Decomposition
• Accuracy of one step of the Falk-Langemeyer method
• Accuracy of the Jacobi method on scaled diagonally dominant symmetric matrices
• Condition numbers and equilibration of matrices
• Convergence of the cyclic and quasi-cyclic block Jacobi methods
• Convergence to diagonal form of block Jacobi-type methods
• Definite matrix polynomials and their linearization by definite pencils
• Detecting a definite Hermitian pair and a hyperbolic or elliptic quadratic eigenvalue problem, and associated nearness problems
• Globally convergent Jacobi methods for positive definite matrix pairs
• scientific article; zbMATH DE number 3164069 (Why is no real title available?)
• scientific article; zbMATH DE number 3671573 (Why is no real title available?)
• scientific article; zbMATH DE number 47363 (Why is no real title available?)
• Jacobi’s Method is More Accurate than QR
• Matrix algorithms. Vol. 2: Eigensystems
• On pairs of almost diagonal matrices
• On Scaled Almost-Diagonal Hermitian Matrix Pairs
• On the Convergence of the Cyclic Jacobi Method for Parallel Block Orderings
• On the Quadratic Convergence of the Falk–Langemeyer Method
• Perturbation bounds for the definite generalized eigenvalue problem

Cited in

(8)

• The high relative accuracy of the HZ method
• On the quadratic convergence of the complex HZ method for the positive definite generalized eigenvalue problem
• Accuracy of one step of the Falk-Langemeyer method
• scientific article; zbMATH DE number 4057146 (Why is no real title available?)
• On the Quadratic Convergence of the Falk–Langemeyer Method
• scientific article; zbMATH DE number 815446 (Why is no real title available?)
• On the convergence of complex Jacobi methods
• Simultaneous diagonalization via congruence of Hermitian matrices: some equivalent conditions and a numerical solution

This page was built for publication: On the complex Falk-Langemeyer method

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2290913)