[Numpy-tickets] [NumPy] #601: function for computing powers of a matrix

NumPy numpy-tickets@scipy....
Sun Feb 24 18:36:15 CST 2008


#601: function for computing powers of a matrix
--------------------------+-------------------------------------------------
 Reporter:  LevGivon      |        Owner:  somebody
     Type:  enhancement   |       Status:  new     
 Priority:  normal        |    Milestone:  1.0.5   
Component:  numpy.linalg  |      Version:  devel   
 Severity:  normal        |   Resolution:          
 Keywords:                |  
--------------------------+-------------------------------------------------
Comment (by charris):

 Powers can be defined for normal matrices, i.e., matrices for which
 A*transpose(A) = transpose(A)*A, but normality itself is subject to
 numerical fuzziness in computational practice. And for normal matrices you
 still need to deal with square roots of negative reals and the fact that
 there can be a multiplicity of roots, and so on and so forth. I think the
 user needs to know what sort of matrices they are dealing with and what
 they intend before they use a function like this.

 This functionality could be part of a package, perhaps in a package for
 symmetric positive indefinite matrices; those matrices are like the non-
 negative reals. But I don't think numpy should try to implement such a
 specialized function.

 However, raising matrices to (small) non-negative integer powers is
 possible and might cover most needs. A special integer power function
 might do the trick.

-- 
Ticket URL: <http://scipy.org/scipy/numpy/ticket/601#comment:7>
NumPy <http://projects.scipy.org/scipy/numpy>
The fundamental package needed for scientific computing with Python.


More information about the Numpy-tickets mailing list