Error Bounds for the Singular Value Decomposition     Next: Further Details: Error Up: Accuracy and Stability Previous: Computing and

Error Bounds for the Singular Value Decomposition

The singular  value decomposition (SVD) of a real m-by-n matrix A is defined as follows. Let r = min(m , n). The the SVD of A is ( in the complex case), where U and V are orthogonal (unitary) matrices and is diagonal, with . The are the singular values of A and the leading r columns of U and of V the left and right singular vectors, respectively. The SVD of a general matrix is computed by xGESVD      (see subsection 2.2.4).

The approximate error bounds for the computed singular values are The approximate error bounds for the computed singular vectors and , which bound the acute angles between the computed singular vectors and true singular vectors and , are These bounds can be computing by the following code fragment.

EPSMCH = SLAMCH( 'E' )
*  Compute singular value decomposition of A
*  The singular values are returned in S
*  The left singular vectors are returned in U
*  The transposed right singular vectors are returned in VT
CALL  SGESVD( 'S', 'S', M, N, A, LDA, S, U, LDU, VT, LDVT,
\$             WORK, LWORK, INFO )
IF( INFO.GT.0 ) THEN
PRINT *,'SGESVD did not converge'
ELSE IF ( MIN(M,N) .GT. 0 ) THEN
SERRBD  = EPSMCH * S(1)
*     Compute reciprocal condition numbers for singular
*      vectors
CALL SDISNA( 'Left', M, N, S, RCONDU, INFO )
CALL SDISNA( 'Right', M, N, S, RCONDV, INFO )
DO 10 I = 1, MIN(M,N)
VERRBD( I ) = EPSMCH*( S(1)/RCONDV( I ) )
UERRBD( I ) = EPSMCH*( S(1)/RCONDU( I ) )
10    CONTINUE
END IF

For example, if and then the singular values, approximate error bounds, and true errors are given below. Tue Nov 29 14:03:33 EST 1994