摘要
传统遗传算法寻优时存在“早熟收敛”、后期搜索效率低，以及难于平衡选择压力和保持种群多样性的问题。用欧式距离定义子代个体C和种群P中个体的相似性，度量个体对种群多样性贡献的大小。将子代和父代的适应度看作随机变量，定义子代和父代的相关系数，用来度量子代从父代所得信息量的多少。将对种群多样性贡献较高且与父代较为相似的个体引入到种群中，既提高种群多样性，又让算法寻优更有依据，防止算法变成随机搜索，从而改进解的质量。这不但同时提高算法求精和求泛的能力，而且较好平衡选择压力和种群多样性。最后将自适应交叉变异概率用于遗传操作，让其随种群个体适应度的变化而变化。仿真实验表明，替代策略和自适应交叉变异概率对平衡选择压力和保持种群多样性效果较好，降低“早熟收敛”的可能性，而且加快进化速度。
In order to resolve the problem that traditional GA is prone to premature, inefficient in the final stage, and difficult to balance between species diversity and selection pressure, the similarities between offspring and individuals of population was defined with the Euclidean distance in this paper, The fitness offspring and parent was considered as a random variable, and the correlation coefficient between offspring and parent was used to measure the amount of information that offspring derived from the parent. Individuals that contributed much diversity to the population and were similar to their parents were inserted into the population. In this way, it not only increases the population diversity, and prevented the algorithm into a random search, but also improved the solution quality. Reforming is improved while improving the refining, therefore selection pressure and population diversity are balanced better. In the end, adaptive crossover and mutation probability was used in genetic algorithm, which made crossover probability and mutation probability change with the change of fitness of individuals. Simulation results show that the replacement strategy presented in this paper is effective in balancing selection pressures and maintaining diversity of population. To some extent, it reduces the possibility of premature convergence and accelerates the speed of evolution.
出处
《重庆航天职业技术学院学报》
2010年第4期46-51,56,共7页
Journal of Chongqing Aerospace Polytechnic
基金
云南省自然科学基金（2009ZC128M）.
关键词
遗传算法
替代策略
种群多样性
交又概率
变异概率
父代子代相关性
Genetic algorithm
replacement strategies
population diversity
crossover probability
muta-tion probability
parent-offspring correlation
