摘要
以常微分方程组的演化建模问题为主要研究对象,设计了分布式异步并行演化算法,并以128台PIII 500微机通过10Mbps的以太网互联而成的机群系统作为模拟实验环境进行了大规模的并行实验,系统地测试了算法中的一些重要的并行控制参数,包括处理机间的连通度、个体的迁移率和迁移代频等对算法性能的影响,得到了一些崭新的实验结果,给出了一些结果分析,特别是对串行算法和并行算法的最好建模结果进行了比较.
A distributed asynchronous parallel evolutionary modeling algorithm for solving the evolutionary modeling problem of system of ordinary differential equations is proposed. Performed on a simulated parallel environment consisting of 128 Pentium III 500 PCs connected by a 10 Mbps Ethernet, large-scale experiments are done to systematically test the effect of some important parallel control parameters, which include the degree of connectivity between processors, the migration rate and the migration interval of individuals, on the performance of the algorithm. Many new experimental results are achieved as well as some analysis and explanations are given. Especially the best modeling results for the sequential algorithm and the parallel one are compared at the end.
出处
《软件学报》
EI
CSCD
北大核心
2003年第3期443-450,共8页
Journal of Software
基金
Supported by the National Natural Science Foundation of China under Grant Nos.70071042
60073043
60133010 (国家自然科学基金)
the Assisting Project of Ministry of Education of China for Backbone Teachers of University and College (国家教育部高等学校骨
关键词
常微分方程组
并行演化建模
实验
遗传程序设计
并行算法
Computer simulation
Differential equations
Evolutionary algorithms
Genetic algorithms
Local area networks
Mathematical models
Personal computers
