Fortran - Schur decomposition via IMSL and Matlab

You are viewing an archived Text-only version of the thread. To view this thread in it's original format and/or if you want to reply to this thread please [click here]

Author

Schur decomposition via IMSL and Matlab

Abdul Qat

2005-08-24, 6:59 pm

I'm trying to reconcile differences between IMSL MP F90's and Matlab's
Schur [U,T] decomposition of a real matrix of moderate size whose spectral
radius and departure from normality are equally large (~10^14) and whose
eigenvalues are observed values lying along the diagonal. In both cases the
norm of the residual in the decomposition is finite (~10^-1).

IMSL's lin_eig_gen correctly computes the eigenvalues. However, they differ
significantly (>> 1ulp) from the diagonal entries in IMSL's T, especially
as they decrease in magnitude. In contrast, Matlab's diagonal entries in T
are indistinguishable from the eigenvalues. In both cases, U is apparently
unitary. Why this difference?

--
Salaam,
Abdul Q.

Sponsored Links