[SciPy-User] scipy.sparse.linalg.eigen error code?

Robert Cimrman cimrman3@ntc.zcu...
Wed Oct 13 12:14:57 CDT 2010


On Wed, 13 Oct 2010, Nico Schlömer wrote:

> Hi,
>
> I'm using scipy.sparse.linalg.eigen to compute the lowest magnitude
> eigenvalue of a matrix, and I noticed that sometimes the code would
> return 0+0j where I really didn't expect it. Turns out that tweaking
> the number of iterations results in some more meaningful value here,
> suggesting that the underlying Arnoldi (ARPACK?) iteration failed with
> the maxiter value previously given. As far as I can see, there's no
> way to tell that the iteration actually failed.
>
> Is that correct?

I have stumbled upon this problem just yesterday, so I can confirm that. 
Sometimes it helped to request more than one eigenvalue, or tweaking the 
ncv parameter.

r.


More information about the SciPy-User mailing list