一种基于灾变的多群体遗传算法

A Catastrophe-based Parallel Genetic Algorithm

在多群体遗传算法的实现中,迁移算子常会产生大量通信开销,也不利于搜索出多个模态峰点。针对上述不足,提出了一种改进''的多群体遗传算法,用灾变产生新个体方式代替子群体之间的迁移效果,同时提出了尖点灾变模型及灾变发生的分歧条件。各子群遗传方式受灾变条件的控制。仿真结果验证了算法的有效性。

Often exist the high communication costs requested in the processes of parallelization of GA because of the migration between the sub-populations, and also the migration operator would destroy the island's characteristic so that the problem of homogeneity between the sub-populations may occur in the convergence phase. Considering the above deficiency, this paper presents a multi-population processing method. Some of new or good genetic materials are created by catastrophe rather than based on the migration .The genetic mechanism of the population is partly dominated by a model called cusp-catastrophe. The catastrophe condition is investigated and applied in the approach. The simulation results show the effectiveness of this approach.