A QR-type reduction for computing the SVD of a general matrix product/quotient

Please use this identifier to cite or link to this item: https://doi.org/10.1007/s00211-002-0431-z

Title:	A QR-type reduction for computing the SVD of a general matrix product/quotient
Authors:	Chu, D. De Lathauwer, L. De Moor, B.
Issue Date:	Jul-2003
Citation:	Chu, D., De Lathauwer, L., De Moor, B. (2003-07). A QR-type reduction for computing the SVD of a general matrix product/quotient. Numerische Mathematik 95 (1) : 101-121. ScholarBank@NUS Repository. https://doi.org/10.1007/s00211-002-0431-z
Abstract:	In this paper, a QR-type reduction technique is developed for the computation of the SVD of a general matrix product/quotient A = A1 s1 A2 s2 ⋯ Am sm with Ai ∈ Rn×n and si = 1 or si = -1. First the matrix A is reduced by at most m QR-factoizations to the form Q11 (1) (Q21 (1))-1, where Q11 (1), Q21 (1) ∈ Rn×n and (Q11 (1)T Q11 (1) + (Q21 (1))T Q21 (1) = I. Then the SVD of A is obtained by computing the CSD (Cosine-Sine Decomposition) of Q11 (1) and Q21 (1) using the Matlab command gsvd. The performance of the proposed method is verified by some numerical examples.
Source Title:	Numerische Mathematik
URI:	http://scholarbank.nus.edu.sg/handle/10635/102740
ISSN:	0029599X
DOI:	10.1007/s00211-002-0431-z
Appears in Collections:	Staff Publications

There are no files associated with this item.

Check