[SciPy-user] Sparse matrix questions

Nils Wagner nwagner at mecha.uni-stuttgart.de
Tue Feb 14 06:48:32 CST 2006


Ed Schofield wrote:
>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? ;)
>
>
>  
Well, my programming skills are not a quarter as good (at least for
integration in scipy).

>># 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
>>>>        
>
>  

  File "/usr/local/lib/python2.4/site-packages/scipy/sparse/sparse.py", line 214, in __getattr__
    raise AttributeError, attr + " not found"
AttributeError: __float__ not found
 

>># 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
>
>_______________________________________________
>SciPy-user mailing list
>SciPy-user at scipy.net
>http://www.scipy.net/mailman/listinfo/scipy-user
>  


 




More information about the SciPy-user mailing list