Hello
How should I modify svdcmp to deal with complex matrices?

Thank you for your answers

Gianluca

SVD definition for complex matrices:
Conjugate[Transpose[u]].DiagonalMatrix[w].v

Read a tread "Complex matrices and svdcmp" from ECC in this forum.

See also http://idlastro.gsfc.nasa.gov/idl_html_help/S77.html.

Evgeny

Hello Evgeny,

Can you direct me please to a good source for complex SVD, assuming I have no background in FORTRAN, and I got somewhat lost trying to figure out the call tree of LAPACK's SVD routine.

I am currently trying to make the necessary modifications using Golub and Van Loan's "Matrix Computations".

Best,
Lior

Lior, Sorry,

I am not working in the field and don't know needed referencies.

I think that there is a few chance to find a free code on C for rectangular complex matrix SVD.

Generally speaking, it is hard to find simple SVD code at level higher than NC (or these commercial C packages, or black box from Fortran, or code translated by f2c, or block matrix linear algebra is used, etc, etc, etc... ).

May be you can avoid using of complex SVD in your physical problem: do matrix inversion by another way or
reformulate problem as SVD for real symmetric matrix:
Conjugate(A).A or
you need only singular values without vectors, etc..?

Evgeny

Lior,

I have found a reference in http://www.cas.mcmaster.ca/~qiao/publications/zsvd2.pdf#search='Complex%20matrix%20SVD%20.c
The article describes SVD of 2x2 complex matrix (C code).

As it is noted in this article, this function may be useful at final stage of SVD of mxn complex matrix after Jacobi method (not Householder).

Evgeny.