基于RDD的分布式粒子群优化算法  

Distributed Particle Swarm Optimization Algorithm Based on Resilient Distributed Datasets

作　　者：程磊生 吴志健[1,2] 彭虎[1,2] 吴双可 邓长寿[3] 王则林[1,2] 

机构地区：[1]武汉大学软件工程国家重点实验室,武汉430072 [2]武汉大学计算机学院,武汉430072 [3]九江学院信息科学与技术学院,江西九江332005

出　　处：《小型微型计算机系统》2016年第11期2542-2546,共5页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61364025;61070008)资助;中央高校基本科研业务费专项资金项目(2012211020205)资助;武汉大学软件工程国家重点实验室开放基金项目(SKLSE2012-09-39)资助;江西省教育厅科学技术项目(GJJ13729)资助;河北省科技支撑计划项目(12210319)资助;南通市科技局应用研究项目(BK2014057)资助

摘　　要：在演化计算领域,粒子群优化算法具有易实现、收敛快、调参少等优点.但是随着问题规模的增大,粒子群优化算法易陷入求解精度不高、耗时过长的窘态,因此本文提出一种基于弹性分布式数据集(Resilient Distributed Datasets,RDD)的分布式粒子群优化算法.该算法采用岛模型将种群分解为若干个岛屿,即子种群,然后使用RDD并行数据结构将整个种群并行化,使得每个岛屿对应RDD中的一个分区,借助RDD的分区并行,实现了粒子群优化算法在分布式平台上的并行.最后,对包括单峰函数和多峰函数在内的11个标准测试函数,将该算法与多种改进的PSO算法进行了比较实验与分析,结果表明该算法求解精度高且加速效果明显.In the field of evolutionary computation, particle swarm optimization algorithm has a certain number of advantages, such as easy realization, fast convergence and less tuning parameters. However, with the size of the problem increasing, particle swarm optimization algorithm gets stuck in the dilemma of low accuracy and time-consuming,thus this paper presents a distributed particle swarm optimization based on resilient distributed datasets. In this algorithm, the global population is divided into several sub-populations, namely islands and uses data structure RDD to make each of islands correspond to one partition. With the partition computing model of RDD, the algorithm has achieved a parallel particle swarm optimization in distributed platform. Finally, this algorithm has been compared with some modified PSO on 11 benchmark functions including unimodal functions and multimodal functions, the result demonstrates that this algorithm has high precision as well as obvious acceleration.

关 键 词：粒子群优化 弹性分布式数据集 岛模型 并行计算 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]