Я бы использовал промежуточную матрицу и метод sum(axis)
:
np.random.seed(24)
# example matrix filled either with 0 or -1:
A = np.random.randint(2, size=(3, 2, 2)) - 1
# condition test:
X_elementwise = A < 0
# Check whether the conditions are fullfilled at least once:
X = X_elementwise.sum(axis=0) >= 1
Значения для A и X:
A = array([[[-1, 0],
[-1, 0]],
[[ 0, 0],
[ 0, -1]],
[[ 0, 0],
[-1, 0]]])
X = array([[ True, False],
[ True, True]])