基因表达式编程中动态适应的远缘繁殖策略
Outbreeding Strategy with Dynamic Fitness in Gene Expression Programming
作者：姜玥(四川大学 计算机学院，四川 成都 610064)；唐常杰(四川大学 计算机学院，四川 成都 610064)；郑明秀(四川大学 计算机学院，四川 成都 610064)
Author：(School of Computer Sci., Sichuan Univ., Chengdu 610064,China)；(School of Computer Sci., Sichuan Univ., Chengdu 610064,China)；(School of Computer Sci., Sichuan Univ., Chengdu 610064,China)
收稿日期：2006-05-09 年卷（期）页码：2007,39(2):121-126
期刊名称：工程科学与技术
Journal Name：Advanced Engineering Sciences
关键字：基因表达式编程；多样性；适应度函数；强变异
Key words：Gene Expression Programming(GEP); diversity; fitness function; strong mutation
基金项目：国家自然科学基金资助项目(60473071);高等学校博士学科点专项科研基金SRFDP资助项目(2002061007)
中文摘要
传统基因表达式编程采用静态适应度函数，影响了后期进化速度和整体质量。提出了远缘繁殖策略和动态适应度函数策略，远亲繁殖并及时变换评估个体的标准，增加多样性并有利于选择优质个体；实验表明，将远缘繁殖和动态适应度函数策略结合，有效地改善了传统GEP的性能，进化代数平均下降达10%，平均最高适应度值提高7%～14%，最高适应度提高达7%以上，进化的成功率提高达30%以上。
英文摘要
Traditional Gene Expression Programming(GEP) uses static fitness function. It decreases evolution speed and quality. To solve the problem, outbreeding strategy and dynamic fitness function strategy: outbreed and timely change standard of evaluating individuals were proposed to increase diversity and select excellent individuals; The experiments showed that the performance of traditional GEP is effectively improved by the combination of OBS and DFFS. The average fitness for the best individual is increased by 7%～14%, the maximal fitness is increased by over 7%, and the success rate is increased by over 30%.
【关闭】
