Einsum multiply each row with every one for 3X3X3 array

Question

Hello could someone please help me figure out how to use np.einsum to produce the below code's result. I have a (3,3,3) tensor and I will like to get this results which I got from using two for loops. the code I wrote to produce this output is below. I am trying to use np.einsum to produce this same result attained from using two for loops in the below code. I am not familar with using np.einsum. Ideally I will also like to sum each of the resulting rows to get nine values.


Command Line Arguments
result of code below   
[1 1 1]
[2 2 2]
[1 1 1]
[2 2 2]
[4 4 4]
[2 2 2]
[1 1 1]
[2 2 2]
[1 1 1]
[1 1 1]

3
6
3
9
12
6
15
18
9
6
12
6
18
24
12

import numpy as np
bb=[]
for x in range(3):
    for y in range(3):
        bb.append((x,y))
a = np.array([[[1,2,1],[3,4,2],[5,6,3]],
             [[1,2,1],[3,4,2],[5,6,3]],
             [[1,2,1],[3,4,2],[5,6,3]]])
b = np.array([[[1,2,1],[3,4,2],[5,6,3]],
             [[1,2,1],[3,4,2],[5,6,3]],
             [[1,2,1],[3,4,2],[5,6,3]]])
for z in range(9):
    llAI  = bb[z]
    aal = a[:,llAI[0],llAI[1]]
    for f in range(9):
        mmAI=bb[f]
        aam = a[:,mmAI[0],mmAI[1]]
        print(np.sum(aal*aam))

hpaulj · Accepted Answer · 2022-12-05T17:12:15.213

It took a bit to figure out what you are doing,

Since z iterates on range(3), aal is successively a[:,0,0], a[:,0,1],a[:,0,2].

Or done all at once:

In [178]: aaL = a[:,0,:]; aaL
Out[178]: 
array([[1, 2, 1],
       [1, 2, 1],
       [1, 2, 1]])

aam does the same iteration. So the sum of their products, using matmul/@/dot is:

In [179]: aaL.T@aaL
Out[179]: 
array([[ 3,  6,  3],
       [ 6, 12,  6],
       [ 3,  6,  3]])

Or in einsum:

In [180]: np.einsum('ji,jk->ik',aaL,aaL)
Out[180]: 
array([[ 3,  6,  3],
       [ 6, 12,  6],
       [ 3,  6,  3]])

Your indexing array:

In [183]: bb
Out[183]: [(0, 0), (0, 1), (0, 2), (1, 0), (1, 1), (1, 2), (2, 0), (2, 1), (2, 2)
In [185]: np.array(bb)[:3,:]
Out[185]: 
array([[0, 0],
       [0, 1],
       [0, 2]])

If I generalize it to the remaining ranges of bb:

In [192]: for i in range(3):
     ...:     aaL = a[:,i]
     ...:     print(aaL.T@aaL)
     ...:     
[[ 3  6  3]
 [ 6 12  6]
 [ 3  6  3]]
[[27 36 18]
 [36 48 24]
 [18 24 12]]
[[ 75  90  45]
 [ 90 108  54]
 [ 45  54  27]]

Adding a dimension to the einsum:

In [195]: np.einsum('jmi,jmk->mik', a,a)
Out[195]: 
array([[[  3,   6,   3],
        [  6,  12,   6],
        [  3,   6,   3]],

       [[ 27,  36,  18],
        [ 36,  48,  24],
        [ 18,  24,  12]],

       [[ 75,  90,  45],
        [ 90, 108,  54],
        [ 45,  54,  27]]])

Hello again I was just wonder what about the case when I have a np array that has a shape of (31,22,16) and I want the resulting shape to be (352,352)? — mike talker, Dec 07 '22 at 05:32
`einsum` does not `reshape`. It can transpose. The connection between those two shapes is not obvious. — hpaulj, Dec 07 '22 at 05:49
Hi again sorry to disturb you again but I think I made a mistake in my question. The following loop should have been this for x in range(3*3): for y in range(3*3): bb.append((x,y)) — mike talker, Dec 08 '22 at 15:04

Einsum multiply each row with every one for 3X3X3 array

1 Answers1