[SciPy-dev] Trouble with optimize.fmin_ncg

[Nils Wagner]

On Wed, 25 Jul 2007 11:33:09 -0700
  "Nathan Bell" <wnbell at gmail.com> wrote:
> On 7/25/07, Nils Wagner <nwagner at iam.uni-stuttgart.de> 
>wrote:
>> Thank you very much! It took me a while to locate
>> the bug. Now it works fine again !
>> BTW, is there a way to support sparse vectors such
>> that the outer product of sparse vectors yields
>> a sparse matrix ?
> 
> If the vector is represented as an Nx1 sparse matrix V, 
>then the outer
> product is simply V*V.T.  Likewise if V is 1xN then the 
>outer product
> is V.T*V.  The existing sparse matrix multiplication 
>code should do
> these operations efficiently.
> 
  
Great !
And how can I assign complex entries ?

from scipy import *
from scipy.sparse import *
n = 10
A = sparse.lil_matrix((n,1),complex)
A[4,0] = 2.
A[-1,0] = -4.+1j
print (A*A.T).todense()

Nils

     A[-1,0] = -4.+1j
   File 
"/usr/local/lib64/python2.5/site-packages/scipy/sparse/sparse.py", 
line 2370, in __setitem__
     x = self.dtype.type(x)
TypeError: can't convert complex to float; use abs(z)