一种基于罚函数的混合分布估计算法

Hybrid estimation of distribution algorithms based on penalty function

分布估计算法是一种新型的基于概率模型的进化计算方法,已在许多领域得到了非常成功的应用.借签罚函数根本思想,把非线性约束优化转变为无约束优化,并利用多变量相关的MIMIC算法对所得的无约束问题进化求解,提出的新算法突破了传统基于约束保持法或可行规则法的约束处理,且分布估计算法是基于可行解的宏观层面的随机进化算法,具有较强全局寻优能力和较高的收敛率.数值试验表明该算法具有很强的全局寻优能力和有效性.

Estimation of distribution algorithms（ EDAs） is a novel probabilistic model based evolutionary algorithm,which has been successfully applied in many areas. Application of EDAs to the nonlinear constrained optimization problems is described in the paper. The constrained problems are transformed to unconstrained problems through penalty function,then MIMIC algorithm of the Multivariate Correlation is applied to solve the unconstrained problems. The proposed algorithm breaks the commonly used method of constraint preserving and feasible rule. In addition,EDAs is macro- level evolutionary approach based on feasible space,with a stronger global search ability and higher convergence. The simulation results show the powerful ability of global searching and efficiency of the method.