[Attempt] Binary array (1,0) sorting in O(1)

最新推荐文章于 2024-10-16 17:48:16 发布

Iyon_

最新推荐文章于 2024-10-16 17:48:16 发布

阅读量233

点赞数

文章标签： numpy python 开发语言

本文链接：https://blog.csdn.net/Iyon_/article/details/125688195

版权

该博客探讨了如何在理论上用O(1)时间复杂度对二进制数组进行排序，尽管实际操作中可能由于字符串转换和数组切片等步骤导致时间消耗并非严格O(1)。首先，通过numpy生成随机二进制数组，然后将数组转换为十进制数以计数数组中的1，最后生成目标排序的数组。然而，测试显示不同长度数组的处理时间并不恒定，可能因字符串转换和数组操作的非线性时间成本导致。

摘要由CSDN通过智能技术生成

[Attempt] Binary array (1,0) sorting in O(1)

Here we try to sort a binary array consisting of zeros and ones. This sorting problem can be also considered as a list summation problem or ones counting problem.

Firstly, we use a uniformly distributed binary array generated by numpy as the raw test array.

# generate random binary list
p = 0.5
size = 10
test_arr = np.random.binomial(1, p, size).astype(np.str0)

Secondly, as a perporcessing step, we convert the list (array) to a decimal number. Unfortunately, this step cannot be guaranteed to be in time complexity O(1).

test_list = list(test_arr)
test_arr_string_int = int(''.join(test_list))

Then, we can count ones in the decimal number by adding the digits. To make the adding process in O(1), the summation is calculated as follows.

digit sum = (num_value-1) % 9 + 1
Leetcode 258.

# count ones in the list
test_arr_ones_count = (test_arr_string_int-1) % 9 + 1

Finally, the array of desired order can be generated by the number of ones.

# zeros after ones
res_arr_1 = np.ones(test_arr.shape)
res_arr_1[test_arr_ones_count:] *= 0
# ones after zeros
res_arr_0 = np.ones(test_arr.shape)
res_arr_0[:test_arr_ones_count] *= 0

Example result

numpy random seed: 0
generate random binary list: 
['1' '1' '1' '1' '0' '1' '0' '1' '1' '0']
convert list to (string) to int:(no guarantee of O(1)
['1', '1', '1', '1', '0', '1', '0', '1', '1', '0']
1111010110
count ones in the list:
7
generate new list of desired order:
[1. 1. 1. 1. 1. 1. 1. 0. 0. 0.]
[0. 0. 0. 0. 0. 0. 0. 1. 1. 1.]

After numbers of test, we found the time consumptions of different length (vaules) of arrays are not exactly the same. This might mean that the whole process is not strictly O(1). This could be due to the following reasons.

String, integer converting process.
Numpy array slicing process.
Instability of the machine.

Complete code

import numpy as np
import time
# np.random.seed(0)
# print('generate random binary list: ')
# generate random binary list
for i in range(100):
    p = 0.5
    size = np.random.randint(100)
    test_arr = np.random.binomial(1, p, size).astype(np.str0)
    # print(test_arr)
    # print('convert list to (string) to int:(no guarantee of O(1)')
    # convert list to (string) to int:(no guarantee of O(1)
    test_list = list(test_arr)
    # print(test_list)
    test_arr_string_int = int(''.join(test_list))
    start = time.time()
    # print(test_arr_string_int)
    # print('count ones in the list:')
    # count ones in the list
    test_arr_ones_count = (test_arr_string_int-1) % 9 + 1
    # print(test_arr_ones_count)
    # print('generate new list of desired order:')
    # zeros after ones
    res_arr_1 = np.ones(test_arr.shape)
    res_arr_1[test_arr_ones_count:] *= 0
    # ones after zeros
    res_arr_0 = np.ones(test_arr.shape)
    res_arr_0[:test_arr_ones_count] *= 0
    # print(res_arr_1)
    # print(res_arr_0)
    print("{} th test, array len:{}, time usage: {}".format(i, size, time.time()-start))