计算机工程与应用 ›› 2016, Vol. 52 ›› Issue (1): 12-16.

• 理论与研发 • 上一篇    下一篇

CUDA平台下信息熵多种群遗传算法设计

李正夫1,2,王希诚3,李克秋1,姚  翔3,董悦丽2   

  1. 1.大连理工大学 计算机科学与技术学院,辽宁 大连 116024
    2.大连东软信息学院 计算机科学与技术系,辽宁 大连 116023
    3.大连理工大学 工业装备结构分析国家重点实验室,辽宁 大连 116024
  • 出版日期:2016-01-01 发布日期:2015-12-30

Information entropy multi-population genetic algorithm based on CUDA

LI Zhengfu1,2, WANG Xicheng3, LI Keqiu1, YAO Xiang3, DONG Yueli2   

  1. 1.School of Computer Science and Technology, Dalian University of Technology, Dalian, Liaoning 116024, China
    2.Department of Computer Science and Technology, Dalian Neusoft University of Information, Dalian, Liaoning 116023, China
    3.State Key Laboratory of Structural Analyses for Industrial Equipment, Dalian University of Technology, Dalian, Liaoning 116024, China
  • Online:2016-01-01 Published:2015-12-30

摘要: 为了进一步提高信息熵多种群遗传算法的计算效率,缩短计算时间,提出了一种基于CUDA平台的信息熵多种群遗传算法。通过分析原算法的并行因素,结合CUDA开发平台,对原算法进行适合GPU加速的并行化处理,实现了遗传算子、惩罚函数和空间收缩因子等的并行计算,有效地提高了算法效率。例题数值测试表明,在保持了快速收敛特性和计算精度的前提下,CUDA并行算法相对于原算法具有很高的加速效率。

null

关键词: 统一计算设备架构(CUDA), 并行计算, 遗传算法, 信息熵, 多种群

Abstract: In order to improve the computational efficiency of information entropy multi-population genetic algorithm, and reduce the computing time, an information entropy multi-population genetic algorithm based on CUDA is proposed. By analyzing the parallelism factors of original algorithm, considering the CUDA platform, parallel processing is made on the original algorithm to suit for GPU-accelerated. Genetic operators, penalty function, and space contraction factors are also modified for CUDA parallelism. All these work improve the efficiency of the original algorithm. Under the premise of keeping the fast convergence and accuracy, example numerical tests show that CUDA parallel algorithms has a high acceleration efficiency.

Key words: Compute Unified Device Architecture(CUDA), parallel algorithm, genetic algorithm, information entropy, multi-population