[SciPy-user] generalized eigenvalue problem for sparse matrices

Ondrej Certik ondrej@certik...
Tue Apr 15 10:08:29 CDT 2008


On Tue, Apr 15, 2008 at 4:55 PM, Matthieu Brucher
<matthieu.brucher@gmail.com> wrote:
>
>
> 2008/4/15, Ondrej Certik <ondrej@certik.cz>:
>
> > On Tue, Apr 15, 2008 at 4:29 PM, Matthieu Brucher
> > <matthieu.brucher@gmail.com> wrote:
> > > Hi,
> > >
> > > I've tested this function with a 2000x2000 array, and like Abhinav, I
> didn't
> > > manage to find correct eigenvalues. Everything worked as expected with a
> > > dense array with the same element with numpy.linalg.eig, but not with
> > > Arpack...
> >
> >
> >
> > When I compiled arpack myself and used it myself, I also didn't manage
> > to get correct results. Anyway, I wouldn't recommend to use arpack.
> >
> >
> > Ondrej
> >
>
> What do you recommend for generalized eigenproblems with sparse arrays ?

I have a very good experience with pysparse. That's why I was
objecting to it's removal from scipy. But I think I got ok to put it
back in, I just didn't find time for that.

Plus there are other good solvers like bzlpack or primme, both should
be able to do generalized eigenvalues and both are more modern than
arpack, but I haven't yet tried them extensively, but it's on my todo
list to create a scikit with easy to use interface to all of them. Any
help appreciated. :)

Ondrej


More information about the SciPy-user mailing list