Numpy 1D-array operation in a ND-array

Question

Let's say we have a 4D-array A of shape (n, m, g, h) and a 1D-array B of shape (n).

So I want to do a power operation for every 3D sub-array of A (m, g, h) with every element of B (which is an int).

A = np.arange(24).reshape(3, 2, 2, 2)
>>>array([[[[ 0,  1],
         [ 2,  3]],

        [[ 4,  5],
         [ 6,  7]]],


       [[[ 8,  9],
         [10, 11]],

        [[12, 13],
         [14, 15]]],


       [[[16, 17],
         [18, 19]],

        [[20, 21],
         [22, 23]]]]
B = np.arange(3)
>>>array([0, 1, 2])

The result that I want is:

C = somefunc(A, B) # just an example, can be anything
>>>array([[[[ (ignore, 0^0),  1],
         [ 1,  1]],

        [[ 1,  1],
         [ 1,  1]]],


       [[[ 8,  9],
         [10, 11]],

        [[12, 13],
         [14, 15]]],


       [[[256, 289],
         [324, 361]],

        [[400, 441],
         [484, 529]]]]

What is the best way to get the desired result?

I thought about something like:

A = [np.power(A[i, :, :, :], B[i]) for i in range(B.size)]

But that would be quite inefficient if B.size is a big number. Any other ideas?

What is inefficient of A = [np.power(A[i, :, :, :], b) for b in B)] ? The a**b is necessary even if b is big. — Albert Alonso
– Albert Alonso, Commented Mar 14, 2019 at 22:59
np.power(A, B[:,None,None,None]) - that is, expand the dimensions of B so it broadcasts with A. — hpaulj
– hpaulj, Commented Mar 14, 2019 at 23:07

fountainhead · Accepted Answer · 2019-03-14 23:41:10Z

1

This will work:

C = A ** B[:,None,None,None]
print (C)

Output:

[[[[  1   1]
   [  1   1]]

  [[  1   1]
   [  1   1]]]


 [[[  8   9]
   [ 10  11]]

  [[ 12  13]
   [ 14  15]]]


 [[[256 289]
   [324 361]]

  [[400 441]
   [484 529]]]]

answered Mar 14, 2019 at 23:41

fountainhead

3,7421 gold badge11 silver badges18 bronze badges

Sign up to request clarification or add additional context in comments.

Collectives™ on Stack Overflow

Numpy 1D-array operation in a ND-array

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related