摘要
遗传算法的本质决定了它的搜索方式是有向随机的，导致了其计算结果具有非稳定性．为了研究遗传算法进化过程中的非稳定性规律，文中首次提出了进化截止代数和进化截止代数分布两个新概念，并给出了它们的具体定义．然后，以工程优化中常用的浮点型遗传算法为例，通过大量的数值试验和统计分析揭示了遗传算法进化截止代数分布的规律．最后，从信息熵的观点出发，用最大信息熵原理对其规律作出了理论上的合理解释．
The essence of genetic algorithms characterizes their searching by the mode of directed random, which causes the result to be unstable. In order to investigate the law of the unstability during the evolutionary process in genetic algorithms, two novel concepts of the truncated generation and the distribution of truncated generations are firstly presented, and their definitions are also detailed. Then, on the basis of considerable amounts of numerical tests and statistical analysis for truncated generations, the distribution law is revealed with an example of float type genetic algorithms utilized frequently in engineering optimization. And finally, from the viewpoint of information theoretic entropy, the law is explained using maximum entropy principle.
出处
《计算机研究与发展》
EI
CSCD
北大核心
2000年第2期188-193,共6页
Journal of Computer Research and Development
关键词
遗传算法
优化
信息熵
代数分布
genetic algorithm, optimization, information theoretic entropy
