[Numpy-discussion] : How to sum weighted matrices

shu wei mailshuwei@gmail....
Wed Mar 9 18:59:30 CST 2011


---------- Forwarded message ----------
From: shu wei <mailshuwei@gmail.com>
Date: Wed, Mar 9, 2011 at 6:34 PM
Subject: Re: [Numpy-discussion] How to sum weighted matrices
To: Nicolas SCHEFFER <scheffer.nicolas@gmail.com>


Thanks very much.
 I just copied the code your guys gave me. Both of them work great
I used to program in Matlab. not familiar with the array multiplication in
python numpy.
I have looked around for some basic explanation on this.
However,I am still very confused with the array axes.

Could someone explain a little bit about these?
Many thanks in advance




On Tue, Mar 8, 2011 at 11:29 AM, Nicolas SCHEFFER <
scheffer.nicolas@gmail.com> wrote:

> Or just with a dot:
>
> ===
> In [17]: np.tensordot(weights, matrices, (0,0))
>
> Out[17]:
> array([[ 5.,  5.,  5.],
>       [ 5.,  5.,  5.]])
>
> In [18]: np.dot(matrices.T,weights).T
>
> Out[18]:
> array([[ 5.,  5.,  5.],
>       [ 5.,  5.,  5.]])
> ==
> make matrices.T C_CONTIGUOUS for maximum speed.
>
> -n
>
>
> On Mon, Mar 7, 2011 at 6:03 PM, shu wei <mailshuwei@gmail.com> wrote:
> > Thanks very much. It works.
> >
> > On Mon, Mar 7, 2011 at 11:53 AM, <qubax@gmx.at> wrote:
> >>
> >> for your problem, you can do:
> >>
> >> ----------------------------
> >>
> >> import numpy as np
> >>
> >> weights = np.array([1,2])
> >>
> >> matrix1 = np.ones((2,3))
> >> matrix2 = 2*np.ones((2,3))
> >>
> >> matrices = np.array([matrix1,matrix2])
> >>
> >> weighted_sum = np.tensordot(weights, matrices, (0,0))
> >>
> >> --------------------------
> >>
> >> On Mon, Mar 07, 2011 at 06:16:15AM -0600, shu wei wrote:
> >> >    Hello all,
> >> >
> >> >    I am new to python and numpy.
> >> >    My question is how to sum up N weighted matrices.
> >> >    For example w=[1,2] (N=2 case)
> >> >    m1=[1 2 3,
> >> >           3 4 5]
> >> >
> >> >    m2=[3 4 5,
> >> >           4 5 6]
> >> >    I want to get a matrix Y=w[1]*m1+w[2]*m2 by using a loop.
> >> >
> >> >    My original problem is like this
> >> >    X=[1 2 3,
> >> >         3 4 5,
> >> >         4 5 6]
> >> >
> >> >    a1=[1 2 3]  1st row of X
> >> >    m1=a1'*a1 a matirx
> >> >    a2=[3 4 5] 2nd row of X
> >> >    m2=a2'*a2
> >> >    a3=[ 4 5 6] 3rd row of X
> >> >    m3=a3'*a3
> >> >
> >> >    I want to get Y1=w[1]*m1+w[2]*m2
> >> >                          Y2=w[1]*m2+w[2]*m3
> >> >    So basically it is rolling and to sum up the weighted matries
> >> >    I have a big X, the rolling window is relatively small.
> >> >
> >> >    I tried to use
> >> >
> >> >    sq=np.array([x[i].reshape(-1,1)*x[i] for i in np.arange(0,len(x)])
> #
> >> >    s=len(x)
> >> >    m=np.array([sq[i:i+t] for i in np.arange(0,s-t+1)]) # t is the
> len(w)
> >> >
> >> >    then I was stuck, I tried to use a loop somethig like
> >> >    Y=np.array([np.sum(w[i]*m[j,i],axis=0) for i in np.arange(0,t)] )
> >> >    Any suggestion is welcome.
> >> >
> >> >    sue
> >>
> >> > _______________________________________________
> >> > NumPy-Discussion mailing list
> >> > NumPy-Discussion@scipy.org
> >> > http://mail.scipy.org/mailman/listinfo/numpy-discussion
> >>
> >>
> >> --
> >> There are two things children should get
> >> from their parents: roots and wings.
> >>
> >> The king who needs to remind his people of his rank, is no king.
> >>
> >> A beggar's mistake harms no one but the beggar. A king's mistake,
> >> however, harms everyone but the king. Too often, the measure of
> >> power lies not in the number who obey your will, but in the number
> >> who suffer your stupidity.
> >> _______________________________________________
> >> NumPy-Discussion mailing list
> >> NumPy-Discussion@scipy.org
> >> http://mail.scipy.org/mailman/listinfo/numpy-discussion
> >
> >
> > _______________________________________________
> > NumPy-Discussion mailing list
> > NumPy-Discussion@scipy.org
> > http://mail.scipy.org/mailman/listinfo/numpy-discussion
> >
> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.scipy.org/pipermail/numpy-discussion/attachments/20110309/cd37646a/attachment.html 


More information about the NumPy-Discussion mailing list