[ublas] Matrix-vector and vector-vector operations

From: Albert Strasheim (fullung_at_[hidden])
Date: 2007-05-30 08:38:08

Hello all

In the code I'm writing I typically have my data in a matrix, with each row 
representing a data "point" and each column a dimension of the data.

Frequently, I want to do things like subtract a vector (like a mean vector) 
from all the rows, or divide all the rows by some vector (like a variance 
vector).

I would like to write:

datamatrix -= meanvec;
datamatrix /= varvec;

and sometimes

datarow -= meanvec;
datarow /= varvec;

Unless I am mistaken, one can't do this with uBLAS as-is?

Are defining extra operators outside of the matrix and vector classes the 
right way add these operations so that I can use them in my code? Or maybe 
there's an existing way to write this kind of thing without having to resort 
to a for-loop each time?

For now, I'd settle for operations that work if matrix.size2() == 
vector.size() and vec1.size() == vec2.size(). At some distant point in the 
future one might want to support broadcasting [1]. If the sizes don't match 
up, some exception is thrown.

Any suggestions on how anybody else addressed this issue would be much 
appreciated.

Cheers,

Albert

[1] http://www.scipy.org/EricsBroadcastingDoc