Convolution of 3D array with 1D kernel in Python

Question

Is it possible to convolve 1D array with 3D array? For example:

A is 8x2x2 matrix that I would like to convolve. Assume A has 2x2 sub-matrices of (A = A7 A6 A5... A0) each sub-matrix is 2x2. B is 5x1 array which contains the scalar weights (B0 B1 B2 B3 B4). What I am trying to do is to convolve B array with the first dimension of A array, which is 8 in this case. I know numpy.convolve is available but it doesn't support multidimensions. To clarify my example:Convolution example

What is the dimension of final output matrix? – Amit Vikram Singh Apr 14 '21 at 09:11 — Amit Vikram Singh, Apr 14 '21 at 09:11
Will be same as A which is 8x2x2. – cyberfue Apr 14 '21 at 11:41 — cyberfue, Apr 14 '21 at 11:41

score 1 · Accepted Answer · answered Apr 14 '21 at 11:47

1

Use:

arr_out = np.apply_along_axis(
                lambda x: np.convolve(x, B.flatten(), mode = 'same'), 
                axis = 0, arr = A)

answered Apr 14 '21 at 11:47

Amit Vikram Singh

2,090
10
20

Convolution of 3D array with 1D kernel in Python

1 Answers1