On Thu, Jul 4, 2013 at 12:54 PM, Daniel J Sebald
<address@hidden> wrote:
On 07/04/2013 02:21 PM, John W. Eaton wrote:
On 07/04/2013 03:03 PM, Daniel J Sebald wrote:
I have no clue why ARPACK has the limitation of providing at most N-2
eigenvalues, but that's the way it works. I'm not going to try to fix
that problem.
they use a shifting strategy to get eigenvalues clustered nearest a given shift, so
if you want all the eigenvalues for a large sparse matrix you can call arpack twice,
once for k/2 eigenvalues at the low end of the spectrum, and again for the k/2 at
the high end. "Large" is greater than 300th order even on my klunky laptop; otherwise
it makes more sense to just call the lapack routines if the user wants a large percentage
of the eigenvalues.