机群系统上基于映射和抽样划分的并行聚类算法

采用映射和抽样划分方法,基于MPI消息传递编程模式,在机群系统上设计与实现一种并行聚类算法.该算法将生物基因序列映射成整数值,采用整数值取代字符串进行聚类,使得聚类过程快速,通过多次抽样一次聚类寻找初值在一定程度上避免了聚类结果陷入局部解的问题,优化了聚类质量.在PC机群系统上对基因序列进行并行聚类的实验结果表明该算法获得了较好的加速和可扩展性.

A Parallel Clustering Algorithm Using Mapping and Sampling-Partitioning on the Cluster Computing Systems

Based on MPI message-passing mode,a parallel clustering algorithm is presented and implemented on cluster computing system by applying mapping and sampling-partitioning techniques.This algorithm maps the given biological gene sequences into the corresponding integer values and then executes clustering in use of the integer values instead of the given to speed up the clustering procedure.To find the initial value by multiple sampling and one clustering can avoid the clustering results that are got into locally optimal solution and generate better clustering quality.The experimental results for the biological gene sequences on the cluster of PC computers show that the presented parallel algorithm obtains good speedup and scalability.