I have a 1-D numpy array a with arbitrary size N, and I have other numpy array b with exactly 2 columns and M rows. I want to build a 2-D array with shape (M, N) where the (i,j) element is the bool value b[:,0][i] <= a[j] <= b[:,1][i]. I want a solution that does not go through a for loop.
Example:
import numpy as np
a = np.array([10, 2, 5])
b = np.array([[ 0, 5],[ 5, 10]])
# Expected array
np.array([[False, True, True], [True, False, True]])
The most direct way is to add an axis and broadcast:
(b[:,0][:, np.newaxis] <= a) & (a <= b[:,1][:, np.newaxis])
You can think of the broadcasting in terms of a nested for-loop. You want an output with shape (M,N), you have a with shape (N,) (for j in range(N):) and b[:,0]/b[:,1] with shape (M,) (for i in range(M):). So you would want to cast b[:,0]/b[:,1] to shape (M,1) for another level to the for-loop.
M x 1 for i in range(M):
N for j in range(N):
-----
M x N answer[i,j] = ...
