基于GPU改进的并行人工蜂群算法  

An Improved Parallel Artificial Bee Colony Algorithm Based on GPU

作　　者：王智广[1] 王文亮[1] 张同举[1] 鲁强[1] 刘伟峰[2,3] 

机构地区：[1]中国石油大学(北京)地球物理与信息工程学院,北京102249 [2]中国石油化工股份有限公司石油勘探开发研究院,北京100083 [3]中国石油化工集团公司多波地震技术重点实验室,北京100083

出　　处：《中南民族大学学报（自然科学版）》2013年第4期86-90,114,共6页Journal of South-Central University for Nationalities：Natural Science Edition

基　　金：国家自然科学基金资助项目(60803159);国家科技重大专项资助项目(2011ZX05005-005-006)

摘　　要：相对于先前的并行人工蜂群算法进行了一些改进,主要采用OpenCL本地内存、并行规约等技术,提出了一种基于图形处理器(GPU)改进的并行人工蜂群算法.该算法将采蜜蜂映射为OpenCL一个工作项,跟随蜂采用右邻域优先的局部选择机制.实验结果表明:文中提出的改进并行人工蜂群算法提高了算法的执行效率,收敛速度得到提升.In this paper, compared with previous parallel artificial bee colony algorithm, we have improved this algorithm in many aspects, such as OpenCL local memory and parallel reduction. An improved parallel artificial bee colony algorithm was proposed based on GPU. The process of the improved parallel artificial bee colony algorithm is that each employee bee mappings to OpenCL work item and onlookers use local selection of right neighborhood. The simulation results show that the improved parallel artificial bee colony algorithm has improved efficiency and the convergence rate is raised.

关 键 词：人工蜂群算法 并行处理 图形处理器 开放计算语言 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]