arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 0907.1369cs.DS
收藏

关于{ \sc平衡分离器}的$O(\sqrt[3]{\log n})$ -近似算法

Towards an $O(\sqrt[3]{\log n})$-Approximation Algorithm for {\sc Balanced Separator}

Manjish Pal

摘要:

{ \sc $c$ -平衡分离器}问题是一个图划分问题,其中给定图$G$,目标是找到最小尺寸的切割,使得切割的两侧至少具有$cn$顶点。在本文中,我们提出了新的进展方向的{ \sc $c$ -平衡分离器}问题。更具体地说,我们提出了一个新的家庭的数学计划,这取决于一个参数$ε> 0$,并延长开创性的工作阿罗拉饶-Vazirani({ \sf ARV})\cite{ARV}表明,多项式时间的可解性所提出的家庭的程序意味着改善的近似因子$O(\log^{{1/3} + ε} n)$从最知名的因素$O(\sqrt{\log n})$由于{ \sf ARV}。实际上,对于$ε= 1/3$,我们得到的程序是{ \sf ARV}提出的SDP。对于$ε< 1/3$,这一系列的程序是不凸的,但可以将它们转换成所谓的\emph{\textbf{concave programs}},其中一个优化凸可行集上的凹函数。凹规划的性质允许应用Hoffman \cite{H81}或Tuy \emph{et al} \cite{TTT85}的技术来以任意精度解决这些问题。但是找到一种方法来解决这些程序在多项式时间内收敛的问题仍然是开放的。我们的结果,虽然有条件的,引入了一个新的家庭的程序,这是更强大的半定规划的近似算法的背景下,因此,它将感兴趣的调查这个家庭都在设计高效的算法和证明硬度结果的方向。

英文摘要:

The {\sc $c$-Balanced Separator} problem is a graph-partitioning problem in which given a graph $G$, one aims to find a cut of minimum size such that both the sides of the cut have at least $cn$ vertices. In this paper, we present new directions of progress in the {\sc $c$-Balanced Separator} problem. More specifically, we propose a new family of mathematical programs, which depends upon a parameter $ε> 0$, and extend the seminal work of Arora-Rao-Vazirani ({\sf ARV}) \cite{ARV} to show that the polynomial time solvability of the proposed family of programs implies an improvement in the approximation factor to $O(\log^{{1/3} + ε} n)$ from the best-known factor of $O(\sqrt{\log n})$ due to {\sf ARV}. In fact, for $ε= 1/3$, the program we get is the SDP proposed by {\sf ARV}. For $ε< 1/3$, this family of programs is not convex but one can transform them into so called \emph{\textbf{concave programs}} in which one optimizes a concave function over a convex feasible set. The properties of concave programs allows one to apply techniques due to Hoffman \cite{H81} or Tuy \emph{et al} \cite{TTT85} to solve such problems with arbitrary accuracy. But the problem of finding of a method to solve these programs that converges in polynomial time still remains open. Our result, although conditional, introduces a new family of programs which is more powerful than semi-definite programming in the context of approximation algorithms and hence it will of interest to investigate this family both in the direction of designing efficient algorithms and proving hardness results.

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。

↑