

SGGSVD(l) ) SGGSVD(l) NAME SGGSVD  compute the generalized singular value decomposition (GSVD) of an MbyN real matrix A and PbyN real matrix B SYNOPSIS SUBROUTINE SGGSVD( JOBU, JOBV, JOBQ, M, N, P, K, L, A, LDA, B, LDB, ALPHA, BETA, U, LDU, V, LDV, Q, LDQ, WORK, IWORK, INFO ) CHARACTER JOBQ, JOBU, JOBV INTEGER INFO, K, L, LDA, LDB, LDQ, LDU, LDV, M, N, P INTEGER IWORK( * ) REAL A( LDA, * ), ALPHA( * ), B( LDB, * ), BETA( * ), Q( LDQ, * ), U( LDU, * ), V( LDV, * ), WORK( * ) PURPOSE SGGSVD computes the generalized singular value decomposition (GSVD) of an MbyN real matrix A and PbyN real matrix B: U'*A*Q = D1*( 0 R ), V'*B*Q = D2*( 0 R ) where U, V and Q are orthogonal matrices, and Z' is the transpose of Z. Let K+L = the effective numerical rank of the matrix (A',B')', then R is a K+LbyK+L nonsingular upper triangular matrix, D1 and D2 are Mby(K+L) and Pby(K+L) "diagonal" matrices and of the following structures, respectively: If MKL >= 0, K L D1 = K ( I 0 ) L ( 0 C ) MKL ( 0 0 ) K L D2 = L ( 0 S ) PL ( 0 0 ) NKL K L ( 0 R ) = K ( 0 R11 R12 ) L ( 0 0 R22 ) where C = diag( ALPHA(K+1), ... , ALPHA(K+L) ), S = diag( BETA(K+1), ... , BETA(K+L) ), C**2 + S**2 = I. R is stored in A(1:K+L,NKL+1:N) on exit. If MKL < 0, K MK K+LM D1 = K ( I 0 0 ) MK ( 0 C 0 ) K MK K+LM D2 = MK ( 0 S 0 ) K+LM ( 0 0 I ) PL ( 0 0 0 ) NKL K MK K+LM ( 0 R ) = K ( 0 R11 R12 R13 ) MK ( 0 0 R22 R23 ) K+LM ( 0 0 0 R33 ) where C = diag( ALPHA(K+1), ... , ALPHA(M) ), S = diag( BETA(K+1), ... , BETA(M) ), C**2 + S**2 = I. (R11 R12 R13 ) is stored in A(1:M, NKL+1:N), and R33 is stored ( 0 R22 R23 ) in B(MK+1:L,N+MKL+1:N) on exit. The routine computes C, S, R, and optionally the orthogonal transformation matrices U, V and Q. In particular, if B is an NbyN nonsingular matrix, then the GSVD of A and B implicitly gives the SVD of A*inv(B): A*inv(B) = U*(D1*inv(D2))*V'. If ( A',B')' has orthonormal columns, then the GSVD of A and B is also equal to the CS decomposition of A and B. Furthermore, the GSVD can be used to derive the solution of the eigenvalue problem: A'*A x = lambda* B'*B x. In some literature, the GSVD of A and B is presented in the form U'*A*X = ( 0 D1 ), V'*B*X = ( 0 D2 ) where U and V are orthogonal and X is nonsingular, D1 and D2 are ``diagonal''. The former GSVD form can be converted to the latter form by taking the nonsingular matrix X as X = Q*( I 0 ) ( 0 inv(R) ). ARGUMENTS JOBU (input) CHARACTER*1 = 'U': Orthogonal matrix U is computed; = 'N': U is not computed. JOBV (input) CHARACTER*1 = 'V': Orthogonal matrix V is computed; = 'N': V is not computed. JOBQ (input) CHARACTER*1 = 'Q': Orthogonal matrix Q is computed; = 'N': Q is not computed. M (input) INTEGER The number of rows of the matrix A. M >= 0. N (input) INTEGER The number of columns of the matrices A and B. N >= 0. P (input) INTEGER The number of rows of the matrix B. P >= 0. K (output) INTEGER L (output) INTEGER On exit, K and L specify the dimension of the subblocks described in the Purpose section. K + L = effective numerical rank of (A',B')'. A (input/output) REAL array, dimension (LDA,N) On entry, the MbyN matrix A. On exit, A contains the triangular matrix R, or part of R. See Purpose for details. LDA (input) INTEGER The leading dimension of the array A. LDA >= max(1,M). B (input/output) REAL array, dimension (LDB,N) On entry, the PbyN matrix B. On exit, B contains the triangular matrix R if M KL < 0. See Purpose for details. LDB (input) INTEGER The leading dimension of the array B. LDA >= max(1,P). ALPHA (output) REAL array, dimension (N) BETA (output) REAL array, dimension (N) On exit, ALPHA and BETA contain the generalized singular value pairs of A and B; ALPHA(1:K) = 1, BETA(1:K) = 0, and if MKL >= 0, ALPHA(K+1:K+L) = C, BETA(K+1:K+L) = S, or if MKL < 0, ALPHA(K+1:M)=C, ALPHA(M+1:K+L)=0 BETA(K+1:M) =S, BETA(M+1:K+L) =1 and ALPHA(K+L+1:N) = 0 BETA(K+L+1:N) = 0 U (output) REAL array, dimension (LDU,M) If JOBU = 'U', U contains the MbyM orthogonal matrix U. If JOBU = 'N', U is not referenced. LDU (input) INTEGER The leading dimension of the array U. LDU >= max(1,M) if JOBU = 'U'; LDU >= 1 oth erwise. V (output) REAL array, dimension (LDV,P) If JOBV = 'V', V contains the PbyP orthogonal matrix V. If JOBV = 'N', V is not referenced. LDV (input) INTEGER The leading dimension of the array V. LDV >= max(1,P) if JOBV = 'V'; LDV >= 1 oth erwise. Q (output) REAL array, dimension (LDQ,N) If JOBQ = 'Q', Q contains the NbyN orthogonal matrix Q. If JOBQ = 'N', Q is not referenced. LDQ (input) INTEGER The leading dimension of the array Q. LDQ >= max(1,N) if JOBQ = 'Q'; LDQ >= 1 oth erwise. WORK (workspace) REAL array, dimension (max(3*N,M,P)+N) IWORK (workspace/output) INTEGER array, dimension (N) On exit, IWORK stores the sorting information. More precisely, the following loop will sort ALPHA for I = K+1, min(M,K+L) swap ALPHA(I) and ALPHA(IWORK(I)) endfor such that ALPHA(1) >= ALPHA(2) >= ... >= ALPHA(N). INFO (output)INTEGER = 0: successful exit < 0: if INFO = i, the ith argument had an illegal value. > 0: if INFO = 1, the Jacobitype procedure failed to converge. For further details, see subroutine STGSJA. PARAMETERS TOLA REAL TOLB REAL TOLA and TOLB are the thresholds to determine the effective rank of (A',B')'. Generally, they are set to TOLA = MAX(M,N)*norm(A)*MACHEPS, TOLB = MAX(P,N)*norm(B)*MACHEPS. The size of TOLA and TOLB may affect the size of back ward errors of the decomposition. Further Details =============== 296 Based on modifications by Ming Gu and Huan Ren, Computer Science Division, University of California at Berkeley, USA LAPACK version 3.0 15 June 2000 SGGSVD(l)