numpy : efficient sum computations

[TG]

Hi there.

I want to do some intensive computations with numpy, and I'm
struggling a bit to find myyyyy wayyyyyy. Here is the problem :

m and d are two matrices :

m.shape = (x,y,a,b)
d.shape = (a,b)

I want to return

i.shape = (x,y)

with

i[x,y] = sum(m[x,y] * d)

I already found that

m[:,:] * d

will give me a matrix of shape (x,y,a,b) containing the products.

Now I want to sum up on axis 2 and 3. If I do :

(m[:,:] * d).sum(axis=3). sum(axis=2)

it seems like I get my result.

I'm wondering : is this syntax leading to efficient computation, or is
there something better ?

thanks

Thomas

[Robert Kern]

TG wrote:

Hi there.

I want to do some intensive computations with numpy, and I'm
struggling a bit to find myyyyy wayyyyyy.

The best place to ask numpy questions is on the numpy mailing list.

http://www.scipy.org/Mailing_Lists

Here is the problem :

m and d are two matrices :

>m.shape = (x,y,a,b)
d.shape = (a,b)

I want to return

>i.shape = (x,y)

with

>i[x,y] = sum(m[x,y] * d)

I already found that

>m[:,:] * d

will give me a matrix of shape (x,y,a,b) containing the products.

Now I want to sum up on axis 2 and 3. If I do :

>(m[:,:] * d).sum(axis=3). sum(axis=2)

it seems like I get my result.

I'm wondering : is this syntax leading to efficient computation, or is
there something better ?

That's about it. There's no need for the [:,:], though.

--
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless enigma
that is made terrible by our own mad attempt to interpret it as though it had
an underlying truth."
-- Umberto Eco