[SciPy-User] Permutation convention for LU decomposition

Jason Grout jason-sage@creativetrax....
Sat Oct 30 20:42:49 CDT 2010


I notice that in Lapack, Matlab, and Mathematica, the LU decomposition 
routine for a matrix A returns a P, L, and U matrices so that:

PA=LU

In scipy, however, the LU decomposition routine gives three matrices so 
that:

A=PLU

(i.e., the P matrix is the inverse of the P matrix returned by the other 
software)

I'm curious why this design decision was made.  Was there precedent with 
other software to make it A=PLU instead of PA=LU?  Is it more natural in 
applications?  I realize it's just a convention.

Thanks,

Jason


More information about the SciPy-User mailing list