一种结合路径增强和Push-Relabel的分布式Mincut/Maxflow算法
A Distributed Mincut/Maxflow Algorithm Combining Path Augmentation and Push-Relabel
摘要:
我们开发了一种新的分布式算法的最小割问题。我们的主要目标是解决大型稀疏问题。假设图的顶点被划分为几个区域,该算法执行区域内的路径扩充和区域之间的推重标记风格的更新。区域之间的交互被认为是昂贵的(区域被一个接一个地加载到内存中或位于网络中的单独机器上)。该算法在扫描中工作-通过所有区域。设$B$是与图的区域间边关联的顶点的集合。我们提出了一个顺序和并行版本的算法,终止于最多$2|B|^2+1$扫描。Delong和Boykov的竞争算法在区域内使用推重新标记更新。在固定划分的情况下,我们证明了该算法有一个严格的$O(n^2)$上界的扫描次数,其中$n$是顶点的数目。我们在计算机视觉中的最大流问题的实例上测试了算法的顺序版本。实验上,新算法所需的扫描次数远低于Delong和Boykov的变体。大型问题(最大到$10^8$顶点和$6\cdot 10^8$边)在大约10次扫描中使用不到1GB的内存解决。
英文摘要:
We develop a novel distributed algorithm for the minimum cut problem. We primarily aim at solving large sparse problems. Assuming vertices of the graph are partitioned into several regions, the algorithm performs path augmentations inside the regions and updates of the push-relabel style between the regions. The interaction between regions is considered expensive (regions are loaded into the memory one-by-one or located on separate machines in a network). The algorithm works in sweeps - passes over all regions. Let $B$ be the set of vertices incident to inter-region edges of the graph. We present a sequential and parallel versions of the algorithm which terminate in at most $2|B|^2+1$ sweeps. The competing algorithm by Delong and Boykov uses push-relabel updates inside regions. In the case of a fixed partition we prove that this algorithm has a tight $O(n^2)$ bound on the number of sweeps, where $n$ is the number of vertices. We tested sequential versions of the algorithms on instances of maxflow problems in computer vision. Experimentally, the number of sweeps required by the new algorithm is much lower than for the Delong and Boykov's variant. Large problems (up to $10^8$ vertices and $6\cdot 10^8$ edges) are solved using under 1GB of memory in about 10 sweeps.
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。