Skip to main content
U.S. Department of Energy
Office of Scientific and Technical Information

A parallel algorithm for computing the singular value decomposition of a matrix

Technical Report ·
OSTI ID:5470126
A parallel algorithm for computing the singular value decomposition will be presented. The basic step of the algorithm is a divide and conquer step based upon a rank one modification of a bidiagonal matrix. Numerical difficulties associated with forming the product of a matrix with its transpose are avoided and numerically stable formuale for obtaining left singular vectors after computing updated right singular vectors are derived. A deflation technique is described which together with a robust root finding method will assure computation of the singular values to full accuracy in the residuals and will also assure the orthogonality of the singular vectors. 16 refs., 2 figs.
Research Organization:
Argonne National Lab., IL (USA)
DOE Contract Number:
W-31109-ENG-38
OSTI ID:
5470126
Report Number(s):
ANL/MCS-TM-102; ON: DE88007059
Country of Publication:
United States
Language:
English