arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2003.00094cs.DS
收藏

一种改进的分布式网络最小割算法

Improved Algorithm for Min-Cuts in Distributed Networks

Mohit Daga

摘要:

在这篇论文中,我们提出了快速确定性算法来发现分布式网络中的小割。寻找网络的最小割点对于确保服务质量和可靠性至关重要。在这篇论文中,我们使用了CONGEST模型,这是一个典型的消息传递模型,用于设计和分析分布式网络中的算法。我们调查各种算法技术的CONGEST模型,并给出了最近的结果,以找到削减的概述。我们还描述了优雅的图论思想,如割空间和循环空间,提供了有用的直觉,我们的工作是建立在。我们的贡献是一个新的快速算法找到小切口。我们的算法依赖于一个新的特征树和削减介绍了这篇论文。我们的算法是建立在几个新的算法的想法,再加上我们的表征树和削减,帮助我们找到所需的最小削减。我们的新技术包括树限制半群函数(TRSF),一种新的素描技术,和分层算法。TRSF是关于生成树定义的,并且基于交换半群。这个简单而强大的技术帮助我们确定性地找到尺寸为1的最小割(桥)和尺寸为2的最小割。我们的草图技术采样一个小的,但相关的顶点集,这是足以找到小的最小切割在某些情况下。我们的分层算法发现在较小的子图中的最小割在不同级别的生成树中的节点为支点,并使用它们来决定在完整的图中的最小割。这很有趣,因为它使我们能够表明,即使对于像寻找min-cut这样的全局属性,也可以以协调的方式利用局部信息。

英文摘要:

In this thesis, we present fast deterministic algorithm to find small cuts in distributed networks. Finding small min-cuts for a network is essential for ensuring the quality of service and reliability. Throughout this thesis, we use the CONGEST model which is a typical message passing model used to design and analyze algorithms in distributed networks. We survey various algorithmic techniques in the CONGEST model and give an overview of the recent results to find cuts. We also describe elegant graph theoretic ideas like cut spaces and cycle spaces that provide useful intuition upon which our work is built. Our contribution is a novel fast algorith to find small cuts. Our algorithm relies on a new characterization of trees and cuts introduced in this thesis. Our algorithm is built upon several new algorithmic ideas that, when coupled with our characterization of trees and cuts, help us to find the required min-cuts. Our novel techniques include a tree restricted semigroup function (TRSF), a novel sketching technique, and a layered algorithm. TRSF is defined with respect to a spanning tree and is based on a commutative semigroup. This simple yet powerful technique helps us to deterministically find min-cuts of size one (bridges) and min-cuts of size two optimally. Our sketching technique samples a small but relevant vertex set which is enough to find small min-cuts in certain cases. Our layered algorithm finds min-cuts in smaller sub-graphs pivoted by nodes at different levels in a spanning tree and uses them to make the decision about the min-cuts in the complete graph. This is interesting because it enables us to show that even for a global property like finding min-cuts, local information can be exploited in a coordinated manner.

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。

↑