摘要
一、引言进化算法(EA)是受自然进化所启发的搜索和优化技术,它包括:遗传算法(GA),进化规划(EP),进化策略(ES),和遗传编程(GP)。 GA是John Holland〔,]等人于1975年在美国密歇根大学应用自然选择过程来解决机器学习问题时提出的,它在于搜索有高的适应值的基因结构;
This paper gives an analysis and comparison for selection mechanism in evolutionary algorithms ,random selection,e. g. roulette wheel selection ;competitive selection,e. g. (μ,λ)selection,q-tournament selection. If the initial fitness function of q-tournament selection is normal distribution, we can give out the behavior of the q-tournament selection. To run in parallel EA,the key is the design and implementation of useful migration mechanism,but decentralized selection with global or local gene pools can solve this problem. If the selection variance is higher ,population size is smaller,the search performance will be poor,to improve performance,one must reduce selection variance or increase population size.
出处
《计算机科学》
CSCD
北大核心
1998年第6期45-48,共4页
Computer Science
关键词
进化算法
选择机制
机器学习
人工智能
Evolutionary algorithms ,Selection density, Loss of diversity
