Avoid Numpy Index For loop

Question

Is there any way to avoid using a second for loop for an operation like this?

for x in range(Size_1):
    for y in range(Size_2):
        k[x,y] = np.sqrt(x+y) - y

Or is there a better way to optimize this? Right now it is incredibly slow for large sizes.

Possible duplicate of [Speed up numpy nested loop](http://stackoverflow.com/questions/23565573/speed-up-numpy-nested-loop) — pinkfloydx33, Feb 09 '17 at 20:17

score 2 · Accepted Answer · answered Feb 09 '17 at 20:12

2

Here's a vectorized solution with broadcasting -

X,Y = np.ogrid[:Size_1,:Size_2]
k_out = np.sqrt(X+Y) - Y

answered Feb 09 '17 at 20:12

Divakar

Paul Panzer · Answer 2 · 2017-02-09T20:28:16.553

0

Supplementing Divakar's solution: If Y and X are not new ranges but some preexisting vectors of numbers, use np.ix_:

Y, X = np.array([[1.3, 3.5, 2], [2.0, -1, 1]])
Y, X = np.ix_(Y, X) # does the same as Y = Y[:, None]; X = X[None, :]
out = np.sqrt(Y+X) - X

edited Feb 09 '17 at 20:28

answered Feb 09 '17 at 20:19

Paul Panzer

2 Answers2