基于图形处理器的并行小波突变协同差分进化算法  

Research on Wavelet Mutation Cooperative Differential Evolution Algorithm Based on Graphic Processing Unit

作　　者：刘剑英[1] 

机构地区：[1]大连职业技术学院信息工程学院

出　　处：《小型微型计算机系统》2014年第3期681-684,共4页Journal of Chinese Computer Systems

基　　金：国家科技支撑计划项目(2009BAH41B05)资助

摘　　要：随着工程应用越来越复杂,待优化参数指数增长,原有的参数优化算法丧失了有效性,基于此提出一种基于图形处理器的并行小波突变协同差分进化算法.该方法首先依据协同进化策略构造协同差分进化框架,将高维问题分解为一些子种群,并同时将总问题分配到每个图形处理单元中,进而在子种群内利用小波突变差分进化算法进行同步并行计算,提高全局持续搜索能力.通过高维标准函数对比测试分析,所提算法相对其他方法具有较好的参数优化能力和计算效率.Focusing on the large-scale parameter optimization, a novel wavelet mutation cooperative differential evolution algorithm based on graphic processing unit （ GPU ） is proposed to optimize the criterion functions. According to the cooperative coevolution strategy, the framework of cooperative differential evolution is constructed, and the whole problem is set into each graphic processing unit, which decomposes high-dimensional problem into some sub-populations, and then differential evolution algorithm based on wavelet mutation is adopted in the sub-populations optimization. Thus, the global continuous search capabilities could be improved. Through high-dimensional standard function tests, compared with other methods, the effectiveness of proposed algorithm for parame- ter optimization is verified.

关 键 词：差分进化 协同进化 高维计算 小波函数 突变 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]