Is there a way to sum up all the columns and rows in a 2d array without using np.sum()?-CodePudding

this is my current code:

M = np.array([[1, 2, 3],
              [4, 5, 6]])

def np_sum_rows(M):  
  rows = []
  for i in range(len(M)):
    rows = M[i, 0:len(M[0])
  return rows.sum()

I want the function to return a vector [6 15]. However, the for loop can only store and return 15. I am positively stumped by this problem and cannot think of any way else but using the for loop.

CodePudding user response：

You can use a list comprehension to sum up the rows in the 2D array and return a list of the sums. Here is one way to do it:

def sum_rows(M):
    return [sum(row) for row in M]

This function takes in a 2D array M and returns a list of the sums of each row in M.

Here is an example of how you can use this function:

M = np.array([[1, 2, 3], [4, 5, 6]])
print(sum_rows(M))  # [6, 15]

CodePudding user response：

You could apply reduce method that comes from functional programming background. It's possible in both Python and numpy-based ways.

In Python you need to apply function of two arguments cumulatively to the items of sequence.

import functools
M = M.tolist() #convert array to list in order to make iteration faster
[reduce(lambda x, y: x y, m) for m in M]
>>> [6, 15]

In numpy you don't need to define a function usually. Universal functions (or ufunc for short) such as add, multiply etc. are used instead.

np.add.reduce(M, axis=1)
>>> array([ 6, 15])

Note that more common way is to apply np.sum on axis 1. So it's strongly adviced to get used to a concept of numpy axes.