numpy 矢量化重采样，如 Pandas DataFrame 重采样

Question

I have an (4, 2000) numpy array and want to resample each column (N=4) for every 5 elements with such as max, min, left, right, which makes its shape as (4, 400).

I can do with Pandas.DataFrame using .resample('5Min').agg(~) or with numpy array and for loop like result = [max(input[i:i+5]) for i in range(0, len(input), 5)] . However, it takes amount of time with large input array since it's not vectorized. Is there any way that I can do with vectorized computation with np array?

Answer 1

Here is another way that uses numpy strides under the hood ( a is your array):

from skimage.util import view_as_blocks
a = view_as_blocks(a, (4,5))

Now, you can use methods/slicing for parameters you want:

#max
a.max(-1)[0].T
#min
a.min(-1)[0].T
#left
a[...,0][0].T
#right
a[...,-1][0].T

example:

a
#[[ 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 27 28 29]
# [30 31 32 33 34 35 36 37 38 39]]

output for max
#[[ 4  9]
# [14 19]
# [24 29]
# [34 39]]