Vectorizing three nested loops - NumPy

Question

I have two arrays x.dim = (N,4) and y.dim = (M, M, 2) and a function f(a, b), which takes K and Ldimensional vectors respectively as arguments. I want to obtain an array res.dim = (N, M, M) such that

for n in range(N):
  for i in range(M):
    for j in range(M):
      res[n, i, j] = f(x[n], y[i, j])

Can't get how to use apply in this case. Thanks in advance for help!

def f(a, b):
    return max(0, 1 - np.sum(np.square(np.divide(np.subtract(b, a[0:2]), a[2:4]))))

this can not work, i, j in range(N) ? int object is not iterable? supposing range(N) is list of ints from 0 to N-1 — Marko Mackic
– Marko Mackic, Commented Jul 19, 2016 at 14:38

Divakar · Accepted Answer · 2016-07-19 15:26:34Z

1

Here's a vectorized approach to work with the listed function using NumPy broadcasting and slicing -

# Slice out relevant cols from x
x_slice1 = x[:,None,None,:2]
x_slice2 = x[:,None,None,2:4]

# Perform operations using those slices to correspond to iterative operations
out = np.maximum(0,1-(np.divide(y-x_slice1,x_slice2)**2).sum(3))

answered Jul 19, 2016 at 15:26

Divakar

222k19 gold badges273 silver badges374 bronze badges

Sign up to request clarification or add additional context in comments.

Collectives™ on Stack Overflow

Vectorizing three nested loops - NumPy

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related