numpy滑点矩阵运算

Question

I'm using numpy to do some matrix operations and I'm unsure on how to achieve a particular operation on the following two matrices:

(a)                       (b)          (c)
[1                     [1 -1 0         [2
 1   (some operation)   1 -1 0      =  -3
 1                      0 -1 0          0
 1]                     0  0 0]         0]

So essentially I want to have an equivalently shaped matrix as (a), but with every entry being the entry in (a), "dotted" with the column in (b) it corresponds to.

I'm currently using the following but it seems rather hacky and only gets me scalars and I feel like there must be a better way:

np.dot(a , np.atleast_2d(b[0].T[0]))[0].sum() = 2
etc...

Any suggestions?

Answer 1

If you initialise your data like this:

b=np.zeros((4,3))
a=np.ones((4,1))
b[0:2,0]=1
b[0:3,1]=-1

You can use dot :

  np.dot(a.T,b)

giving

array([[ 2., -3.,  0.]])

Almost what you want. Your problem is that you have no 4th column in b corresponding to the 4th element in a . You could append a 0 to the result or append a column of zeros to b to get around this. Or:

c=np.zeros_like(a)
c[0:3]=np.dot(a.T,b).T