[英]Quicker Method to Group Numpy Array Elements Based on Second Numpy Array
There are 2 NumPy arrays groups
and selectors
, where有 2 个 NumPy 数组groups
和selectors
,其中
selectors
is an array containing integers that needs to be grouped selectors
是一个包含需要分组的整数的数组import numpy as np
np.random.seed(0)
selectors = np.random.randint(0, 300, 5)
# [172 47 117 192 251]
groups
is a structured array containing the first index (int) of a group (str) groups
是包含组 (str) 的第一个索引 (int) 的结构化数组# Generate groups `a` to `t` and their first index
start = ord('a')
groups = []
for i in range(20):
e = (i*i, chr(start+i))
groups.append(e)
groups = np.array(groups, dtype=[('index', np.uint32), ('selector', '|U1')])
groups = np.sort(groups, order='index')
# [( 0, 'a') ( 1, 'b') ( 4, 'c') ( 9, 'd') ( 16, 'e') ( 25, 'f')
# ( 36, 'g') ( 49, 'h') ( 64, 'i') ( 81, 'j') (100, 'k') (121, 'l')
# (144, 'm') (169, 'n') (196, 'o') (225, 'p') (256, 'q') (289, 'r')
# (324, 's') (361, 't')]
Given these example arrays, the desired result after grouping will be a dictionary of np.ndarrays
/lists鉴于这些示例数组,分组后所需的结果将是np.ndarrays
/lists 的字典
{
"g": [47] ,
"k": [117],
"n": [172, 192],
"p": [251]
}
Is there a quicker way to perform this grouping in Numpy instead of nesting 2 loops, as shown below?有没有更快的方法在 Numpy 中执行此分组而不是嵌套 2 个循环,如下所示? This will be useful for large selectors
arrays with 10-100 million rows using groups
array with 100-1000 rows.这对于使用具有 100-1000 行的groups
数组的具有 10-1 亿行的大型selectors
数组非常有用。
Using Nested Loops使用嵌套循环
results = {}
for s in selectors:
for i in range(len(groups)-1):
if s >= groups[i][0] and s < groups[i+1][0]:
j = i
break
else:
j = i + 1
try:
results[groups[j][1]].append(s)
except KeyError:
results[groups[j][1]] = [s]
print(results)
# {'n': [172, 192], 'g': [47], 'k': [117], 'p': [251]}
If you use binary search on each selector, you are effectively changing the time of your routine from O(len(groups) * len(selectors))
to O(log2(len(groups)) * len(selectors))
如果您在每个选择器上使用二分搜索,您实际上将例程的时间从O(len(groups) * len(selectors))
更改为O(log2(len(groups)) * len(selectors))
The Python documentation on the bisect
module explains how to use it to find the right-most element less than or equal to a specified value. bisect
模块的Python 文档解释了如何使用它来查找小于或等于指定值的最右侧元素。
声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.