Numerical methods for general and structured eigenvalue problems. (Q2484097): Difference between revisions
From MaRDI portal
Changed an Item |
Set profile property. |
||
Property / MaRDI profile type | |||
Property / MaRDI profile type: MaRDI publication profile / rank | |||
Normal rank |
Revision as of 08:20, 5 March 2024
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Numerical methods for general and structured eigenvalue problems. |
scientific article |
Statements
Numerical methods for general and structured eigenvalue problems. (English)
0 references
2 August 2005
0 references
This book deals with the numerical methods for general and structured eigenvalue problems. It starts off with the well known QR algorithm, which is a numerically back stable method for computing all eigenvalues and invariant subspaces of a non symmetric matrix. Given a matrix \(A,\) it generates a sequence which converges to a nontrivial block Schur form. By setting \(A_{1}=A\) and \(A_{i+1}=R_{i}Q_{i}\) if \(A_{i}=Q_{i}R_{i}\) is the QR decomposition of \(A_{i}\), we obtain \(A_{i+1}=Q_{i}^{T}A_{i}Q_{i}.\) Chapter one contains classical results on the perturbation analysis of eigenvalues, local and global convergence and related issues of the round-off error. Methods such as balancing, block algorithms, and advanced deflation techniques are also explored. In chapter two, the QZ algorithm for a pair of matrices \((A,B)\) and aggressive early deflation method are exposed. The third chapter focuses on Krylov subspace type methods and variants of the Arnoldi method while the fourth on how these algorithms apply to structured matrices. The book has also two appendices, (A) background in control theory and (B) on software and practical computing. These lecture notes provide a detailed treatment and an updated account on the various eigenvalue algorithms and related topics such as computing in finite precision, fast algorithms, error estimates, CPU time, accuracy, and efficiency gains for structured matrices. In today's world of computing, it is certainly a valuable reference and tool for anyone using methods of numerical linear algebra.
0 references
eigenvalues
0 references
QR algorithm
0 references
textbook
0 references
convergence
0 references
round-off error
0 references
block algorithms
0 references
deflation techniques
0 references
QZ algorithm
0 references
Krylov subspace type methods
0 references
Arnoldi method
0 references
error estimates
0 references