Stable double LR algorithm and its error analysis
A normative matrix \(A\) is a real tridiagonal matrix with unit super- diagonal, sub-diagonal elements \(b_ i\), \(i=2,\dots,n\), and diagonal elements \(a_ i\), \(i=1,\dots,n\). The symmetric tridiagonal matrix \(S\) with the same diagonal and off-diagonal elements \(\beta_ i\), \(i=2,\dots,n\), has the same eigenvalues as \(A\) if \(b_ i= \beta^ 2_ i\), \(i=2,\dots,n\). The author shows that the double LR transformation of \(A\) with origin shift is related to the QR transformation of \(S\) with the same origin shift. The double LR has operation count only 4/7 of that for the QR method. A stable version of the double LR algorithm is then given, followed by the result that a practical version of this is numerically stable.
- The result of two steps of the \(LR\) algorithm is diagonally similar to the result of one step of the \(HR\) algorithm
- QRT: A QR-Based Tridiagonalization Algorithm for Nonsymmetric Matrices
- Convergence of LR algorithm for a one-point spectrum tridiagonal matrix
- The Relation Between the QR and LR Algorithms
This page was built for publication: Stable double LR algorithm and its error analysis
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1333873)
