Question

I'm working with 3D matrices in numpy. I'm actually passing these matrices to C using ctypes to carry out some calculation and then getting back a result. Now the thing is, my result is correct (I did the math on paper to verify), but it's just not in a form I want it to be.

Here's an example. I have a 3D array of the form:

array([[[ 0,  1,  2],
        [ 3,  4,  5],
        [ 6,  7,  8]],

       [[ 9, 10, 11],
        [12, 13, 14],
        [15, 16, 17]],

       [[18, 19, 20],
        [21, 22, 23],
        [24, 25, 26]]])

I need to convert it to a form where the ith columns of all 2D sub-matrices form a new 2D sub-matrix, as so:

array([[[ 0, 9,  18],
        [ 3, 12, 21],
        [ 6, 15, 24]],

       [[ 1, 10, 19],
        [ 4, 13, 22],
        [ 7, 16, 25]],

       [[2, 11, 20],
        [5, 14, 23],
        [8, 17, 26]]])

I have tried using various combinations of np.rot90, np.flipud, np.fliplr, all to no avail. Any help on this would be greatly appreciated.

Thanks a lot!

Answer 1

Your desired output is your initial array with the order of the axes reversed. That's how NumPy generalizes transposes to arbitrary-dimensional arrays, so you can use the T attribute for this:

In [3]: x
Out[3]: 
array([[[ 0,  1,  2],
        [ 3,  4,  5],
        [ 6,  7,  8]],

       [[ 9, 10, 11],
        [12, 13, 14],
        [15, 16, 17]],

       [[18, 19, 20],
        [21, 22, 23],
        [24, 25, 26]]])

In [4]: x.T
Out[4]: 
array([[[ 0,  9, 18],
        [ 3, 12, 21],
        [ 6, 15, 24]],

       [[ 1, 10, 19],
        [ 4, 13, 22],
        [ 7, 16, 25]],

       [[ 2, 11, 20],
        [ 5, 14, 23],
        [ 8, 17, 26]]])

Rearranging 3D numpy arrays in a specific manner

1 个答案: