CODEWITHSUNDEEP
pythonnumpy
Yulian
Yulian

Reputation: 365

Dot product over the second dimension in Numpy?

I have an array of N vectors, each with a size of 3:

[[ 0  1  2]
 [ 3  4  5]
 [ 6  7  8]
 [ 9 10 11]
 [12 13 14]]

It is needed to find a dot product of each of the vectors with itself.
Thus, the result will be a 3-dimensional array of shape (N,3,3). One approach is to use the following for loop:

for vector in np.arange(15).reshape(-1,3):
    np.outer(vector, vector)

Since the array can be arbitrarily large, I need to find a vectorized solution.

Upvotes: 0

Views: 98

Answers (1)

hpaulj
hpaulj

Reputation: 231385

You aren't summing over any axis, are you? Just a 'batched' outer?

In [115]: arr=np.arange(15).reshape(5,3)                                        
In [116]: arr[:,:,None]*arr[:,None,:]     #using broadcasting                                          
Out[116]: 
array([[[  0,   0,   0],
        [  0,   1,   2],
        [  0,   2,   4]],

       [[  9,  12,  15],
        [ 12,  16,  20],
        [ 15,  20,  25]],

       [[ 36,  42,  48],
        [ 42,  49,  56],
        [ 48,  56,  64]],

       [[ 81,  90,  99],
        [ 90, 100, 110],
        [ 99, 110, 121]],

       [[144, 156, 168],
        [156, 169, 182],
        [168, 182, 196]]])
In [117]: _.shape                                                               
Out[117]: (5, 3, 3)

arr[:,:,None]@arr[:,None,:] does the same thing, summing on that size 1 'dummy' dimension.

Upvotes: 2

Related Questions