"Less/greater than" comparison of N-dimensional and (N-k)-dimensional numpy arrays

Question

Given two arrays a=np.array([[1, 3], [3, 4]]) and b=np.array([2, 2]).

The goal: get array np.array([False, True]) by operation like a>b. I.e. compare rows (True if each pair of elements satisfy the > operator else False ) instead of elementwise comparison (i.e. I don't want get np.array([[False, True], [True, True]])).

And similar for 3-D and (optional) N-dimensional arrays. E.g.

a1 = np.array([[[1, 2, 1], [2, 3, 2]], [[3, 4, 3], [4, 3, 4]]])
b1 = np.array([1, 1, 1])

Operation like a1 > b1 have to return np.array([[False, True], [True, True]]).

How to do it?

Answer 1

Solution is found: use additionally numpy.all function.

Usage for my examples:

a=np.array([[1, 3], [3, 4]])
b=np.array([2, 2])
numpy.all(a > b, axis=1)

Result:

array([False,  True], dtype=bool)

And

a1 = np.array([[[1, 2, 1], [2, 3, 2]], [[3, 4, 3], [4, 3, 4]]])
b1 = np.array([1, 1, 1])
numpy.all(a1 > b1, axis=2)

Result:

array([[False,  True],
       [ True,  True]], dtype=bool)

numpy.all also allows pass multiple axes (as tuple of ints), so it can be used for any dimensions.

Also numpy allows to use ndarray.all method of numpy arrays. Then examples can be rewritten as (a>b).all(axis=1) and (a1>b1).all(axis=2) respectively.