[SciPy-dev] lobpcg eigenvalue solver

[Nils Wagner]

Robert Cimrman wrote:
> Hi,
>
> a pure SciPy implementation of Locally Optimal Block Preconditioned
> Conjugate Gradient Method (LOBPCG) for symmetric eigenvalue problems, see
> http://www-math.cudenver.edu/~aknyazev/software/BLOPEX/,
> was added into scipy.sandbox. It was written in collaboration with
> Andrew Knyazev, one of LOBPCG authors, the license is BSD.
>
> LOBPCG solves eigenproblems Ax = lambda Bx with large, possibly sparse,
> symmetric matrix A and symmetric positive definite matrix B.
>
> Matrices could be given as full numpy arrays, sparse scipy matrices, or
>  as functions performing the matrix-vector product.
>
> There is not much documentation yet, so look at the web page above, and
> send me your comments!
>
> r.
> PS: Later I would like to move the module into main scipy tree - where
> would be the best place?
> _______________________________________________
> Scipy-dev mailing list
> Scipy-dev at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-dev
>   
Hi Robert,

Great work ! Thank you very much !
I have just enabled the package. Here is the output of my short test

Traceback (most recent call last):

  File "test_lobpcg.py", line 17, in ?
    w = lobpcg.lobpcg(X,A,B)
  File
"/usr/lib64/python2.4/site-packages/scipy/sandbox/lobpcg/lobpcg.py",
line 148, in lobpcg
    residualTolerance = sqrt( 1e-15 ) * n
NameError: global name 'sqrt' is not defined

Cheers,
                   Nils

P.S. Is there a specific reason why you have used a pure python
implementation ?