[SciPy-user] Sparse matrix questions

Ed Schofield schofield at ftw.at
Tue Feb 14 06:33:21 CST 2006


Nils Wagner wrote:

>from scipy.sparse import *
>from scipy import *
>
>A = rand(4,4)    
>sparseA = csc_matrix(A)
>#
># How can I build the norm of a sparse matrix without using todense ?
>#
>eps = linalg.norm(sparseA.todense())
>  
>
This is a non-trivial problem, and I don't think we have a function for
it.  Could you write one and contribute it? ;)


># How do I scale a sparse matrix object ?
>#
>#sparseA = sparseA/0.5 # doesn't work
>
>  
>
Interesting.  I'll fix this in SVN.  For now, just use multiplication
instead:
>>> sparseA *= 2.0

># How do I initialize a sparse matrix of order n \times m with zeros
>#
>B = zeros((4,5),Float)
>#sparseB = csc_matrix(B) # doesn't work
>  
>
It's filled with zeros already.  Just use
>>> sparseB = csc_matrix((4,5))

But for efficiency I suggest you use dok_matrix to construct and
manipulate matrices, then convert to csc or csr for solvers or matrix
multiplication.

For diagonals, you can use dok_matrix.setdiag():

>>> sparseB = dok_matrix((5,5))
>>> sparseB.setdiag([1]*5)
>>> sparseB.todense()
array([[ 1.,  0.,  0.,  0.,  0.],
       [ 0.,  1.,  0.,  0.,  0.],
       [ 0.,  0.,  1.,  0.,  0.],
       [ 0.,  0.,  0.,  1.,  0.],
       [ 0.,  0.,  0.,  0.,  1.]])


I hope this helps.  In the future, please post questions like these on
the scipy-user list, not on scipy-dev :)

-- Ed



More information about the SciPy-user mailing list