[SciPy-user] Sparse eigensolver

[Nils Wagner]

On Mon, 10 Dec 2007 10:53:27 +0100
  Robert Cimrman <cimrman3 at ntc.zcu.cz> wrote:
> 
> There is also scipy.sandbox.lobpcg (pure scipy code, 
>depending 
> temporarily on symeig, until symeig functionality is in 
>scipy)
> 
> It seems to run with recent scipy, but 
>lobpcg/tests/benchmark.py gives 
> two additional zero eigenvalues that are not in symeig 
>results. Nils, 
> have you tried it recently? This used to work...
  
python -i 
local/lib64/python2.3/site-packages/scipy/sandbox/lobpcg/tests/benchmark.py
****** 128

Results by LOBPCG

128 [   1.    4.    9.   16.   25.   36.   49.   64.   81. 
 100.]

Results by symeig

128 [  1.   4.   9.  16.  25.  36.  49.  64.  81. 100.]
****** 256

Results by LOBPCG

256 [   1.    4.    9.   16.   25.   36.   49.   64.   81. 
 100.]

Results by symeig

256 [  1.   4.   9.  16.  25.  36.  49.  64.  81. 100.]
****** 512

Results by LOBPCG

512 [  1.   4.   9.  16.  25.  36.  49.  64.  81. 100.]

Results by symeig

512 [   1.    4.    9.   16.   25.   36.   49.   64.   81. 
 100.]
****** 1024

Results by LOBPCG

1024 [   1.    4.    9.   16.   25.   36.   49.   64. 
  81.  100.]

Results by symeig

1024 [  1.   4.   9.  16.  25.  36.  49.  64.  81. 100.]
****** 2048

Results by LOBPCG

2048 [  1.   4.   9.  16.  25.  36.  49.  64.  81. 100.]

Results by symeig

2048 [   1.    4.    9.   16.   25.   36.   49.   64. 
  81.  100.]
  
Cheers,

              Nils