Decomposition of an updated correlation matrix via hyperbolic transformation.
Given a rectangular, in general complex, \(m\times n\) matrix \(X\), the problem is to compute singular values of \(X\) and of its update \(X^{\text{new}}\) obtained by omission of certain columns---\(m\times p\) matrix \(Y\)---and addition of new ones--- \(m\times k\)-matrix \(Z\). The classical way is to use the singular decomposition (SVD) algorithm based on the lower triangularization of \(X,\;Y\) and \(Z\). In the paper, another method is presented for computing singular values of \(X^{\text{new}}\), namely the hyperbolic SVD based on the hyperbolic Givens and Householder transformations matrices. Their properties have been proved in [\textit{D. Janovská} and \textit{G. Opfer}, Numer. Linear Algebra 8, 127--146 (2001; Zbl 1051.65045)] and are shortly reviewed here. Then the existence conditions of the hyperbolic SVD are recalled and an algorithm proposed using sequentially the Householder and Givens transformations. Finally, some applications of the algorithm are mentioned.
- scientific article; zbMATH DE number 16399
- On the QR algorithm and updating the SVD and the URV decomposition in parallel
- On the efficient update of the singular value decomposition
- A note on the hyperbolic singular value decomposition without hyperexchange matrices
- Some results on the problem of updating the hyperbolic matrix factorizations
This page was built for publication: Decomposition of an updated correlation matrix via hyperbolic transformation.
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q558537)