[with spkg] sparse eigenvalues and splines
|Reported by:||jkantor||Owned by:||was|
|Report Upstream:||Work issues:|
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).
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)