arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1010.4535math.OCcs.CCmath.CO
收藏

用组合流算法代替谱技术求扩束比、归一化割和电导

Replacing spectral techniques for expander ratio, normalized cut and conductance by combinatorial flow algorithms

Dorit S. Hochbaum

摘要:

聚类、分割和成像中的一些具有挑战性的问题传统上是使用“谱技术”来解决的。这些问题包括归一化割问题、图扩展比问题、Cheeger常数问题和电导问题。这些问题有几个共同的特点:都寻求一组元素的二分性;问题被公式化为一种形式的比割;这里所示的离散最优化公式等价于离散变量的二次比,有时称为罗利比,以及我们称之为平衡或正交性约束的单个和约束;当不考虑变量的离散性时,用谱方法求解连续松弛。事实上,谱松弛技术是一种主要方法,为这些问题提供了近似解。我们提出了一个算法,这些问题只涉及放松的正交约束。对于一个有$n$节点和$m$边的图,这种松弛在强多项式时间内以O(mn log((n^2)/ m)的最优方式求解.该算法以HPF(Hochbaum's Pseudo-Flow)为子程序,能在小于10分钟内解决数百万个元素和3亿多条边的划分问题。它也表明,通过初步的实验研究,提出的组合算法的结果往往显着提高质量的谱方法的结果。

英文摘要:

Several challenging problem in clustering, partitioning and imaging have traditionally been solved using the "spectral technique". These problems include the normalized cut problem, the graph expander ratio problem, the Cheeger constant problem and the conductance problem. These problems share several common features: all seek a bipartition of a set of elements; the problems are formulated as a form of ratio cut; the formulation as discrete optimization is shown here to be equivalent to a quadratic ratio, sometimes referred to as the Raleigh ratio, on discrete variables and a single sum constraint which we call the balance or orthogonality constraint; when the discrete nature of the variables is disregarded, the continuous relaxation is solved by the spectral method. Indeed the spectral relaxation technique is a dominant method providing an approximate solution to these problems. We propose an algorithm for these problems which involves a relaxation of the orthogonality constraint only. This relaxation is shown here to be solved optimally, and in strongly polynomial time, in O(mn log((n^2) / m) for a graph on $n$ nodes and $m$ edges. The algorithm, using HPF (Hochbaum's Pseudo-Flow) as subroutine, is efficient enough to be used to solve these bi-partitioning problems on millions of elements and more than 300 million edges within less than 10 minutes. It is also demonstrated, via a preliminary experimental study, that the results of the combinatorial algorithm proposed often improve dramatically on the quality of the results of the spectral method.

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。

↑