numpy - Divide columns in array without loops

Question

For example:

Input:

a = array([[1, 2], [4, 10], [4, 6]]])

Output:

b = array([[4, 5], [1, 0.6]])

a can have more than three rows.

Answer 1

You can achieve this using np.roll, by dividing a by a-shifted:

>>> a = np.array([[1,2], [4,10], [4,6]])
array([[ 1,  2],
       [ 4, 10],
       [ 4,  6]])

>>> shifted_a = np.roll(a, shift=1, axis=0)
array([[ 4,  6],
       [ 1,  2],
       [ 4, 10]])

>>> shifted_a / a
array([[0.25      , 0.33333333],
       [4.        , 5.        ],
       [1.        , 0.6       ]])

Then discard the first rows:

>>> (shifted_a / a)[1:]
array([[4. , 5. ],
       [1. , 0.6]])