Python-NumPy切片每行不同

Question

I have a small problem, as I am learning Python. I am trying to slice a 2D-array in a particular way : taking one item over two, but at each line we begin at a different index, for example if we have a = np.reshape(np.arange(16),(4,4)) , so

>>print(a) = [[ 0  1  2  3]
              [ 4  5  6  7]
              [ 8  9 10 11]
              [12 13 14 15]] 

We would like to end with

>>print(new_a) = [[ 0  2]
                  [ 5  7]
                  [ 8 10]
                  [13 15]]

I am sure that it is not too complicated, but I couldn't find the answer :( (I now how to slice a np-array, just not how to change each row)

Thank you and have a nice day !

Answer 1

As long as the dimensions of a are even, one can get this specific (checkerboard) pattern with np.einsum :

>>> np.einsum('jiki->jik', a.reshape(2, 2, 2, 2)).reshape(4, 2)
array([[ 0,  2],
       [ 5,  7],
       [ 8, 10],
       [13, 15]])

or, more generally

>>> a = np.arange(40).reshape(4, 10)
>>> np.einsum('jiki->jik', a.reshape(a.shape[0]//2, 2, -1, 2)).reshape(a.shape[0], -1)
array([[ 0,  2,  4,  6,  8],
       [11, 13, 15, 17, 19],
       [20, 22, 24, 26, 28],
       [31, 33, 35, 37, 39]])

Answer 2

You can use slicing with steps . Do this from 0 and from 1 with step size 2. This will give you two separate arrays, you can then do the work and join them back.

I am pretty new to numpy so there may be a better approach.

Answer 3

b = a.take([0,2,5,7,8,10,13,15]).reshape(4,2)
print(b)

array([[ 0,  2],
       [ 5,  7],
       [ 8, 10],
       [13, 15]])