摘要
探讨了避开事先了解和设定任何小生境相关参数的小生境方法。考虑NGA进化的特点,利用最小欧氏距离下的成对个体具有性态相似性及其大概率地同属同一小生境的特点,采用改进的进化算子建立(1+1)模拟自然小生境内性态相似个体的竞争机制,各个小生境内多对性态相似个体的竞争能够保证各小生境的同时进化,同时引入整体解空间的动态交叉和变异概率来保持群体的多样性,构造了一种全新的小生境算法。4个多峰函数优化数值试验结果证明此方法稳定、显效。
A new method is proposed which can avoid predetermining parameters that related to determinate niche. We make the individuals in each population in pairs according to the least Euclidean between two individuals, and consider them （a pair of individuals ）being the similar characters and belonging to the same niche at most probability. （1＋1） competition condition to imitate competition between the individuals being similar characters among nature niches is constructed by improved evolutionary operator. The competitions of some pairs of individuals among niches and using dynamic crossover and mutation probability during evolutionary of genetic algorithm can guarantee evolutionary of each niche and get population diversification at the same time. 4 experimental results show this new NGA is successful and stable.
出处
《计算机与数字工程》
2008年第11期1-3,14,共4页
Computer & Digital Engineering
基金
内蒙古自然科学基金资助(编号:200711020713)项目
内蒙古科技大学校内科研项目基金资助(编号:00810211)项目
关键词
小生境遗传算法
欧氏距离
(1+1)竞争
动态参数
Niche genetic algorithm, Euclidean distance, （1＋1） competition strategy, dynamic character
