Repeat Numpy array by a sliding window

Question

From the following array of shape (6, 3):

>>> arr
[
    [1, 0, 1],
    [0, 0, 2],
    [1, 2, 0],
    [0, 1, 3],
    [2, 2, 1],
    [2, 0, 2]
]

I'd like to repeat the values according to a sliding window of n=4, giving a new array of shape (6-n-1, n, 3):

>>> new_arr
[
    [
        [1, 0, 1],
        [0, 0, 2],
        [1, 2, 0],
        [0, 1, 3]
    ],
    [
        [0, 0, 2],
        [1, 2, 0],
        [0, 1, 3],
        [2, 2, 1]
    ],
    [
        [1, 2, 0],
        [0, 1, 3],
        [2, 2, 1],
        [2, 0, 2]
    ]
]

It is relatively straightforward using a loop, but it gets extremely slow with several million values (instead of 6 in this example) in the initial array.

Is there a faster way to get to new_arr using Numpy primitives?

Riccardo Bucco · Accepted Answer · 2021-11-25 15:00:54Z

3

You can use NumPy, specifically this function (only NumPy >= 1.20.0):

from numpy.lib.stride_tricks import sliding_window_view

new_arr = sliding_window_view(arr, (n, arr.shape[1])).squeeze()

answered Nov 25, 2021 at 15:00

Riccardo Bucco

15.5k4 gold badges29 silver badges57 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

Jivan Over a year ago

Is there any way to achieve this using Numpy 1.19 ? For some reason Numpy >= 1.20 has dependency conflicts with TensorFlow.

mathfux Over a year ago

sliding_window_view is a more user friendly substitute for numpy.lib.stride_tricks.as_strided introduced in numpy 1.20. Just use it with an extreme care.

Collectives™ on Stack Overflow

Repeat Numpy array by a sliding window

1 Answer 1

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related