arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 0811.4376cs.DScs.CC
收藏

快速排序的平均复杂度有多高?

How robust is quicksort average complexity?

Suman Kumar Sourabh, Soubhik Chakraborty

摘要:

本文对目前流行的快速排序算法平均情况时间复杂度的鲁棒性提出了质疑。在本文研究的六个标准概率分布中,只有连续均匀分布、指数分布和标准正态分布支持它,而其他分布都支持最坏情况的复杂性度量。对于这个问题-为什么我们每次都得到最坏情况下的复杂性度量,而平均情况下的度量是不可信的?--一个逻辑答案是在普遍分布下的平均情况复杂度等于最坏情况复杂度。这个答案很难被质疑,但它并没有说明哪种标准概率分布属于普适性的范畴。士气是,平均情况下的复杂性措施,在他们是不同于那些在最坏的情况下,应被视为稳健的,只要他们得到至少从标准的概率分布,离散和连续的支持。遗憾的是,快速排序的情况并非如此。

英文摘要:

The paper questions the robustness of average case time complexity of the fast and popular quicksort algorithm. Among the six standard probability distributions examined in the paper, only continuous uniform, exponential and standard normal are supporting it whereas the others are supporting the worst case complexity measure. To the question -why are we getting the worst case complexity measure each time the average case measure is discredited? -- one logical answer is average case complexity under the universal distribution equals worst case complexity. This answer, which is hard to challenge, however gives no idea as to which of the standard probability distributions come under the umbrella of universality. The morale is that average case complexity measures, in cases where they are different from those in worst case, should be deemed as robust provided only they get the support from at least the standard probability distributions, both discrete and continuous. Regretfully, this is not the case with quicksort.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑