Unexpected collapse of dimension when calling np.tile()

May 25, 2023

I am creating a multi-dimension numpy matrix like this:

a = np.array([255, 0])                            
mins_and_maxes = np.tile(a, [9, 2, 43])

I’m expecting mins_and_maxes to be a 4-D array with a shape of (9, 2, 43, 2). However, mins_and_maxes has a shape of (9, 2, 86). The [255, 0] arrays are sort of being ‘dissolved’. (I can’t think of a better word. "Exploded"?)

How do I get a matrix of size (9, 2, 43) where every element is a copy of the array of length 2, [255, 0]?

>Solution :

You can try:

a = np.array([255, 0])

mins_and_maxes = np.tile(a, [9, 2, 43, 1])

mins_and_maxes.shape
#(9, 2, 43, 2)

mins_and_maxes

#array([[[[255,   0],
         [255,   0],
         [255,   0],
         ...,
         [255,   0],
         [255,   0],
         [255,   0]],
       [[[255,   0],
         [255,   0],
         [255,   0],
         ...,
         [255,   0],
         [255,   0],
         [255,   0]],

        [[255,   0],
         [255,   0],
         [255,   0],
         ...,
         [255,   0],
         [255,   0],
         [255,   0]]],


       [[[255,   0],
         [255,   0],
         [255,   0],
         ...,
         [255,   0],
         [255,   0],
         [255,   0]],

        [[255,   0],
         [255,   0],
         [255,   0],
         ...,
         [255,   0],
         [255,   0],
         [255,   0]]]])