[with spkg] sparse eigenvalues and splines

[jkantor]

scipy has a sandbox directory with experimental or code in development. One of the packages is an arpack wrapper which computes eigenvalues of sparse matrices. matlab uses this in its eigs command. Although the arpack wrapper is not completely finished, it has useful functionality, and since this has been requested multiple times I think including it is good, it will also maybe spur some people to take the wrapper the last mile (me probably).

Also included is a wrapper of fitpack which has some spline functions (bivariate spline, etc).

 http://sage.math.washington.edu/home/jkantor/spkgs/scipy_sandbox.spkg

The following shows how to compute the smallest five eigenvalues (in magnitude) of a 2000x2000 matrix

import arpack
from scipy import sparse
m=sparse.csc_matrix(random_matrix(RDF,2000,2000,density=.02).numpy())
e,v=arpack.eigen(m,5)

[jkantor]

patch with doctests  http://sage.math.washington.edu/home/jkantor/spkgs/arpack_doctest.patch

[mabshoff]

Merged in 2.9.alpha6.

[jkantor]

new patch, makes the doctest more stable (hopefully)  http://sage.math.washington.edu/home/jkantor/spkgs/arpack_doctest_2.patch

(apply after applying the first one)

[mabshoff]

I apply arpack_doctest_2.patch to 2.9.rc0.

Cheers,

Michael