Re: Partial Schur decomposition in Matlab


In article <fg4j75$83c$1@xxxxxxxxxxxxxxx>,
Dirk Diggler <dirk_f_diggler@xxxxxxxxx> writes:
Hi,
What would be the easiest way to compute partial Schur decomposition in
Matlab? Pearhaps modifying eigs.m in such a way that it returns Schur
instead of Ritz vectors? I must admitt I am having significant trouble with
that. Any other sugesstions/solutions?

this will hardly work since eigs is only a calling sequence for arpack
and modyfying arpack yourself something like rewriting it.
achieving this piecewise (one vector obtained, compute Schur transformation,
repeat) is also hopeless since this destroys sparsity.
you seem , in principle, on the way to rediscover G.W.Stewart's SRR algorithm
for simultaneous vector iteration for nonhermitian matrices (published in
Numerische Mathematik 25, pages 123-136, (1976 )
hth
peter
.

Quantcast