如何基于切片实现无for循环的NumPy数组3元素滑动窗口求和？

Hey there! Let's break down why your current code is throwing a scalar error, then walk through a few clean, loop-free ways to get that 3-element sliding window sum you need.

What's Wrong With Your Original Code? #

Your approach of creating start and stop arrays makes sense on paper, but NumPy doesn't allow using arrays as slice bounds in arr[start:stop]. Slice syntax in NumPy expects scalar values for the start/end indices, so passing arrays here causes the scalar mismatch error you're seeing.

Loop-Free Solutions #

Here are three solid methods to achieve your sliding window sum without any for loops:

1. Using sliding_window_view (Most Intuitive)

NumPy 1.20+ has a built-in function specifically for creating sliding window views of arrays. This is the cleanest approach:

import numpy as np

n = 3
arr = np.array([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])

# Create a view of sliding 3-element windows
sliding_windows = np.lib.stride_tricks.sliding_window_view(arr, window_size=n)
# Sum each window along the second axis
sum_arr = sliding_windows.sum(axis=1)

print(sum_arr)
# Output: [ 6  9 12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63 66 69 72]

This creates a view of the original array (no extra memory overhead) and sums each window in a vectorized operation—super efficient.

2. Using Convolution (np.convolve)

Sliding window sum is equivalent to convolving your array with a kernel of ones. This method is concise:

sum_arr = np.convolve(arr, np.ones(n, dtype=int), mode='valid')
print(sum_arr)

The mode='valid' parameter ensures we only keep results where the kernel fully overlaps with the array, which matches your requirement of consecutive 3-element sums.

3. Manual Shift & Add (Great for Understanding)

If you want to see the mechanics behind the sliding window, you can shift the array three times and add the corresponding elements:

sum_arr = arr[:-2] + arr[1:-1] + arr[2:]
print(sum_arr)

Here, arr[:-2] takes elements 0-22, arr[1:-1] takes 1-23, and arr[2:] takes 2-24. Adding them together gives the sum of each consecutive trio.

All three methods will give you the same correct result, and none use loops. Pick the one that fits your readability needs and NumPy version!

内容的提问来源于stack exchange，提问作者tony selcuk