How to find the sum of every 4 elements in an array

user_15587323 0
Viewed 348

I have an array that will print a 10x10 random numbers from 1 to 9. I want to sum every 4 elements horizontally and vertically.

Here is the code:

import numpy as np

random_array = np.random.randint(1, 10, size=(10, 10))
print(random_array)

For example:

[[7 4 5 7 3 9 4 2 5 3]
 [2 3 7 1 1 1 4 3 2 8]
 [8 2 6 8 8 1 7 8 8 3]
 [4 7 8 8 9 5 9 2 9 7]
 [2 1 9 2 4 5 4 6 5 2]
 [3 3 1 6 4 4 7 9 7 9]
 [2 8 3 2 7 8 7 3 9 2]
 [4 4 6 5 4 1 7 5 1 1]
 [9 5 8 1 9 9 9 1 1 4]
 [8 8 6 1 3 6 4 1 8 3]]

What I want to do is add every 4 elements horizontally and vertically. E.g. horizontally a sum of (7, 4, 5, 7) and then the sum of (4, 5, 7, 3) and so on. Then vertically the sum of(7,2,8,4) and sum of (2,8,4,2) and so on.

numpy arrays python
2 Answers

You can use slicing and np.sum

>>> a = np.random.randint(1,10, size=(10,10))
>>> a
array([[2, 5, 5, 6, 2, 7, 4, 4, 2, 7],
       [1, 4, 4, 5, 7, 1, 3, 4, 3, 5],
       [7, 8, 3, 4, 9, 7, 2, 9, 6, 8],
       [8, 8, 2, 7, 4, 3, 4, 2, 5, 6],
       [9, 7, 3, 1, 8, 4, 8, 6, 4, 3],
       [7, 9, 1, 6, 9, 9, 1, 6, 7, 5],
       [7, 6, 6, 7, 1, 4, 3, 5, 1, 9],
       [5, 4, 4, 2, 3, 8, 9, 2, 7, 2],
       [8, 2, 4, 2, 2, 8, 6, 9, 6, 7],
       [9, 4, 6, 3, 4, 2, 1, 1, 5, 2]])

>>> for i in range(4,11): 
...     print(a[0,i-4:i]) #horizontal
...
[2 5 5 6]
[5 5 6 2]
[5 6 2 7]
[6 2 7 4]
[2 7 4 4]
[7 4 4 2]
[4 4 2 7]

>>> for i in range(4,11):
...     print(a[i-4:i,0]) # vertical
...
[2 1 7 8]
[1 7 8 9]
[7 8 9 7]
[8 9 7 7]
[9 7 7 5]
[7 7 5 8]
[7 5 8 9]

## summing horizontally

>>> for i in range(4,11):
...     s+= np.sum(a[0,i-4:i])
...
>>> s
126
user_116225160

I would do this with convolution using np.convolve — it will be faster than looping, and I think it's more elegant than fiddling about with slices.

import numpy as np

np.random.seed(42)  # For reproducibility.
random_array = np.random.randint(1, 10, size=(10,10))

kernel = np.ones((1, 4))  # <-- 'Template' to sum a group of this shape.

# Horizontal sums.
sum_h = convolve2d(random_array, kernel, mode='valid')
print(sum_h)

This produces:

[[24. 24. 23. 22. 25. 23. 24.]
 [25. 22. 16. 21. 21. 18. 21.]
 [17. 19. 20. 15. 23. 23. 21.]
 [24. 28. 23. 22. 22. 17. 24.]
 [18. 21. 22. 19. 16. 10. 15.]
 [18. 18. 22. 18. 14. 14. 16.]
 [29. 24. 20. 20. 24. 31. 27.]
 [33. 25. 26. 25. 20. 20. 20.]
 [12. 16. 22. 28. 32. 33. 26.]
 [23. 27. 23. 24. 22. 20. 18.]]

For vertical sums you would transpose the kernel so that it's a column of ones:

convolve2d(random_array, kernel.T, mode='valid')

Specifying mode='valid' ensures that you only get the convolution result where the template fits completely over the input matrix.

user_33813050
Related