An Algorithm for $Ax = \lambda Bx$ with Symmetric and Positive-Definite <i>A</i> and <i>B</i> (Q3980609)

From MaRDI portal

!

This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:

scientific article; zbMATH DE number 22702
Language Label Description Also known as
default for all languages
No label defined
    English
    An Algorithm for $Ax = \lambda Bx$ with Symmetric and Positive-Definite <i>A</i> and <i>B</i>
    scientific article; zbMATH DE number 22702

      Statements

      An Algorithm for $Ax = \lambda Bx$ with Symmetric and Positive-Definite <i>A</i> and <i>B</i> (English)
      0 references
      0 references
      0 references
      26 June 1992
      0 references
      algorithm
      0 references
      generalized eigenvalue problem
      0 references
      Cholesky decompositions
      0 references
      generalized singular value problem
      0 references
      \(CS\) decomposition
      0 references
      \(QR\) decomposition
      0 references
      \(QZ\) algorithm
      0 references
      numerical examples
      0 references
      An algorithm for computing the solution of the generalized eigenvalue problem \(Ax=\lambda Bx\) with real symmetric and positive \(n\times n\) matrices \(A\) and \(B\) is considered. By means of the Cholesky decompositions \(A=LL^ T\) and \(B=L_ BL^ T_ B\) the problem is reduced to the generalized singular value problem \(LL^ Tx= \lambda(L_ BL^ T_ B)x\). The problem of computing the generalized singular value decomposition is reduced to the computation of the \(CS\) decomposition of the matrix \(Q\) from the \(QR\) decomposition of \((L,L_ B)^ T=QR\). Finally, \(A\) and \(B\) are reduced to diagonal form by the singular value decomposition. The provided algorithm is faster than the \(QZ\) algorithm: if only the eigenvalues are desired then this algorithm requires \(3h^ 3\) flops in contrast to the \(QZ\) algorithm with \(15h^ 3\) flops. The algorithm is also stable. Numerical examples are presented.
      0 references

      Identifiers