基于多尺度量子谐振子算法的相空间概率聚类算法  

Phase space probabilistic clustering algorithm based on multi-scale quantum harmonic oscillator algorithm

作　　者：王梓懿[1] 安俊秀[1] 王鹏[2] 

机构地区：[1]成都信息工程大学并行计算实验室,成都610225 [2]西南民族大学计算机科学与技术学院,成都610225

出　　处：《计算机应用》2017年第8期2218-2222,共5页journal of Computer Applications

基　　金：国家自然科学基金资助项目(71673032)~~

摘　　要：针对大型集群难以进行任务调度和资源分配的问题,提出一种基于多尺度量子谐振子算法的相空间概率聚类算法(PSPCA-MQHOA)。首先,将集群工作状态投影到相空间中,把复杂的集群工作状态转化为相空间中的点集;进而,将相空间网格化,形成多尺度量子谐振子算法(MQHOA)以处理离散目标函数;最后,利用MQHOA优化过程中波函数变化的概率解释对集群节点进行概率聚类。PSPCA-MQHOA继承了MQHOA物理模型明确、搜索能力强、结果精确等优点,并且由于以相空间作为离散化的目标函数,迭代次数大大减少。实验结果表明PSPCA-MQHOA能适用于多种负载状态的集群。A Phase Space Probabilistic Clustering Algorithm based on Multi-scale Quantum Harmonic Oscillator Algorithm（ PSPCA-MQHOA） was proposed to solve the task scheduling and resource allocation of large clusters. Firstly, the cluster operating status was projected into the phase space, and the complex working state was transformed into the point set in the phase space. Furthermore, the phase space was meshed to form the Multi-scale Quantum Harmonic Oscillator Algorithm（ MQHOA） for discrete objective function. Finally, probabilistic clustering of cluster nodes was carried out by using the probability interpretation of wave function in the MQHOA process. PSPCA-MQHOA inherits the advantages of MQHOA, such as explicit physical model, strong search capabilities and accurate results, and it has few iterations due to the discretized phase space. Experimental results show that PSPCA-MQHOA can be applied to clusters in a variety of load conditions.

关 键 词：概率聚类 量子谐振子 相空间 波函数 集群 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]