我需要一个滚动窗口(又名滑动窗口)可迭代的序列/迭代器/生成器。(默认的Python迭代可以被认为是一种特殊情况，其中窗口长度为1。)我目前正在使用以下代码。我怎样才能做得更优雅和/或更有效?
def rolling_window(seq, window_size):
it = iter(seq)
win = [it.next() for cnt in xrange(window_size)] # First window
yield win
for e in it: # Subsequent windows
win[:-1] = win[1:]
win[-1] = e
yield win
if __name__=="__main__":
for w in rolling_window(xrange(6), 3):
print w
"""Example output:
[0, 1, 2]
[1, 2, 3]
[2, 3, 4]
[3, 4, 5]
"""
对于window_size == 2的特定情况(即，在序列中迭代相邻的重叠对)，请参见如何从列表中迭代重叠(当前，下一个)值对?
修改了DiPaolo的答案，允许任意填充和可变步长
import itertools
def window(seq, n=2,step=1,fill=None,keep=0):
"Returns a sliding window (of width n) over data from the iterable"
" s -> (s0,s1,...s[n-1]), (s1,s2,...,sn), ... "
it = iter(seq)
result = tuple(itertools.islice(it, n))
if len(result) == n:
yield result
while True:
# for elem in it:
elem = tuple( next(it, fill) for _ in range(step))
result = result[step:] + elem
if elem[-1] is fill:
if keep:
yield result
break
yield result
深度学习中滑动窗口数据的优化函数
def SlidingWindow(X, window_length, stride):
indexer = np.arange(window_length)[None, :] + stride*np.arange(int(len(X)/stride)-window_length+4)[:, None]
return X.take(indexer)
应用于多维数组
import numpy as np
def SlidingWindow(X, window_length, stride1):
stride= X.shape[1]*stride1
window_length = window_length*X.shape[1]
indexer = np.arange(window_length)[None, :] + stride1*np.arange(int(len(X)/stride1)-window_length-1)[:, None]
return X.take(indexer)
这似乎是为collections.deque定制的，因为您实际上有一个FIFO(添加到一端，从另一端删除)。然而，即使你使用列表，你也不应该切片两次;相反，您应该只从列表中弹出(0)并追加()新项。
下面是一个基于deque的优化实现:
from collections import deque
def window(seq, n=2):
it = iter(seq)
win = deque((next(it, None) for _ in xrange(n)), maxlen=n)
yield win
append = win.append
for e in it:
append(e)
yield win
在我的测试中，它在大多数时候都轻松击败了这里发布的其他所有东西，尽管pillmuncher的tee版本在大可迭代对象和小窗口方面击败了它。在较大的窗口上，deque再次以原始速度领先。
Access to individual items in the deque may be faster or slower than with lists or tuples. (Items near the beginning are faster, or items near the end if you use a negative index.) I put a sum(w) in the body of my loop; this plays to the deque's strength (iterating from one item to the next is fast, so this loop ran a a full 20% faster than the next fastest method, pillmuncher's). When I changed it to individually look up and add items in a window of ten, the tables turned and the tee method was 20% faster. I was able to recover some speed by using negative indexes for the last five terms in the addition, but tee was still a little faster. Overall I would estimate that either one is plenty fast for most uses and if you need a little more performance, profile and pick the one that works best.
我的两个版本的窗口实现
from typing import Sized, Iterable
def window(seq: Sized, n: int, strid: int = 1, drop_last: bool = False):
for i in range(0, len(seq), strid):
res = seq[i:i + n]
if drop_last and len(res) < n:
break
yield res
def window2(seq: Iterable, n: int, strid: int = 1, drop_last: bool = False):
it = iter(seq)
result = []
step = 0
for i, ele in enumerate(it):
result.append(ele)
result = result[-n:]
if len(result) == n:
if step % strid == 0:
yield result
step += 1
if not drop_last:
yield result
为什么不
def pairwise(iterable):
"s -> (s0,s1), (s1,s2), (s2, s3), ..."
a, b = tee(iterable)
next(b, None)
return zip(a, b)
它被记录在Python文档中。
您可以轻松地将其扩展到更宽的窗口。
