基于浮点数编码的信息熵控制多种群遗传算法

Multi-population Genetic Algorithm Controlled by Information Entropy Based on Floating-point Coding

在用准精确惩罚函数处理约束优化问题的基础上 ,提出一种基于浮点数编码机制的信息熵控制多种群遗传算法。通过在遗传设计中定义一个新的概率而引入信息熵概念 ,构造出一个信息熵优化模型。该模型不必完全求解 ,即可容易求出作为概率的拉格朗日乘子 ,得出空间收缩概率 ,控制各种群中解空间的收缩。信息熵的介入可使优化过程更加平稳 ,收敛更快。同时 ,该算法给出了一种科学而有效的遗传设计收敛判据。实例证明该文算法在求解约束优化问题时快速。

An improved floating-point coded genetic algorithm controlled by information entropy is presented to solve the constrained optimization problems based on the quasi-exact penalty function.The concept of information entropy is introduced into the genetic evolution by defining the probability that the optimal solution located in each population,then a multi-objective model including information entropy is constructed.By the use of this model,the probability can be straightly obtained subsequently,the coefficient of the designed space of variables narrowing down for each population can be got to control the populations searching the optimal solution.The introduction of information entropy makes the optimization procedure more stable and the convergence speed faster.Besides, a new scientific and efficient convergent rule is used in this paper. Numerical examples are given to demonstrate the efficiency of the proposed algorithm.