Doing a certain numpy multiplication row-wise

Question

I have two NumPy arrays that I would like to multiply with each other across every row. To illustrate what I mean I have put the code below:

import numpy as np 

a = np.array([
     [1,2],
     [3,4],
     [5,6],
     [7,8]])


b = np.array([
     [1,2],
     [4,4],
     [5,5],
     [7,10]])

final_product=[]
for i in range(0,b.shape[0]):
    product=a[i,:]*b
    final_product.append(product)

Rather than using loops and lists, is there more direct, faster and elegant way of doing the above row-wise multiplication in NumPy?

score 2 · Accepted Answer · edited Sep 03 '22 at 02:13

2

By using proper reshaping and repetition you can achieve what you are looking for, here is a simple implementation:

a.reshape(4,1,2) * ([b]*4)

If the length is dynamic you can do this:

a.reshape(a.shape[0],1,a.shape[1]) * ([b]*a.shape[0])

Note : Make sure a.shape[1] and b.shape[1] remains equal, while a.shape[0] and b.shape[0] can differ.

edited Sep 03 '22 at 02:13

Ali_Sh

2,667
3
43
66

answered Sep 02 '22 at 22:17

skt7

1,197
8
21

score 2 · Answer 2 · answered Sep 03 '22 at 02:09

2

This type of problems can be handled by np.einsum(see Doc & this post) for more understanding. It is one of the most efficient ways in this regard:

np.einsum("ij, kj->ikj", a, b)

answered Sep 03 '22 at 02:09

Ali_Sh

2,667
3
43
66

score 1 · Answer 3 · answered Sep 02 '22 at 22:14

Try:

n = b.shape[0]
print(np.multiply(np.repeat(a, n, axis=0).reshape((a.shape[0], n, -1)), b))

Prints:

[[[ 1  4]
  [ 4  8]
  [ 5 10]
  [ 7 20]]

 [[ 3  8]
  [12 16]
  [15 20]
  [21 40]]

 [[ 5 12]
  [20 24]
  [25 30]
  [35 60]]

 [[ 7 16]
  [28 32]
  [35 40]
  [49 80]]]

Doing a certain numpy multiplication row-wise

3 Answers3