CODEWITHSUNDEEP
pythonnumpy
piccolo
piccolo

Reputation: 2217

Concise way to sum selected rows of a numpy array

I have a 2D numpy array L, which I want to convert into another numpy array of the same shape such that each row is replaced by the sum of all the other rows. I have demonstrated this below.

My question is if there is a more concise/elegant way of doing this (preferably using more advanced numpy syntax/tools).

L = np.array([[ 0,  1,  2],
              [ 3,  4,  5],
              [ 6,  7,  8],
              [ 9, 10, 11]])

store = []
for i in range(L.shape[0]):
    store.append(np.sum(L,axis=0) - L[i])
output = np.stack(store)

Which gives me the correct output:

array([[18, 21, 24],
       [15, 18, 21],
       [12, 15, 18],
       [ 9, 12, 15]])

Upvotes: 1

Views: 976

Answers (1)

Divakar
Divakar

Reputation: 221574

Simply subtract L from the column-summations and hence leverage broadcasting too in the process for a vectorized solution -

In [12]: L.sum(0) - L
Out[12]: 
array([[18, 21, 24],
       [15, 18, 21],
       [12, 15, 18],
       [ 9, 12, 15]])

Upvotes: 2

Related Questions