numpy array, strangeness

[Alex Martelli]

Stephane Jolicoeur wrote:

> Hello,
> I am trying to work with LinearAlgebra.py, it works but Numpy won't let
> me do the reverse equation.
> I have an array q that is 12 by 9 then use LinearAlgebra's
> singular_value_decomposition on it and get back 3 arrays T , S and D.
> Now q should be the product of T*S*D, so I do this in the shell and get

Whoa!  In Numeric, T*S*D means *ELEMENT-WISE* multiplication of
the three arrays T, S, and D -- and their shapes are such that
such element-wise multiplication fails (and if it succeeded you
would not be happy about it!!!).

What you need is MATRIX multiplication, which in Numeric is
performed by function Numeric.matrixmultiply, also known as
Numeric.dot.


Alex