多点交叉学习组织进化算法

为了提高组织进化算法(organizational evolutionary algorithm,简称OEA)在高维多模函数全局优化中陷入局部极值,分析了OEA 算法早熟收敛的原因,提出了多点交叉学习组织进化算法(mOEA).设计了一个多个组织的领导交叉学习策略来提高组织领导种群多样性,避免早熟收敛;结合社会群体认知和学习的习性,改进OEA 中的吞并算子,使得同一组织内的个体成员有的在其领导周围爬山运动,有的在搜索范围内随机变异,既提高成员群体的适应度值,又增强成员群体的多样性,不易陷入局部极值.与OEA

Multipoint Crossover Learning Organizationary Evolutionary Algorithm

The original organizational evolutionary algorithm (OEA) is often trapped in local optima when optimizing multimodal functions with high dimensions. In this paper, following an analysis of the main causes of the premature convergence, it proposes a novel algorithm, called the multipoint organizational evolutionary algorithm (mOEA). To discourage the premature convergence, a crossover strategy of multiple points is designed to achieve a better diversity of leader population. Inspired by the cognition and learning physics of social swarms, an improved annexing operator enables members in an organization to either partially climb around their leader or randomly mutate within the search range. The new annexing manipulation both enhances fitness values and preserves a good diversity of member population. Experiments on six complex optimization benchmark functions with 30 or 100 dimensions and very large numbers of local minima show that, comparing with the original OEA and CLPSO, mOEA effectively converges faster, results in better optima, is more robust.