多峰搜索的自适应遗传算法

Adaptive genetic algorithm for multi-peak searching

对多峰函数问题提出了基于峰值转换和优育子群相结合的遗传搜索策略.主要是:通过变换函数将多峰问题中的所有峰变成“等高”峰,从而保证每个峰都有同等机会被找到;在种群中实施各种遗传操作及近亲排斥策略,以保证种群的多样性;将种群中适应值超过阈值的个体迁徙形成一个子群,在子群中实施“梯度操作”,对个体进行精细进化.该方法不仅可保证较快地找到所有峰,而且无需对多峰函数做峰的个数已知、峰均匀健分布等任何先验假设.最后与Spears的简单子群法进行了对比实验.

An adaptive multi-peak genetic searching strategy based on optimal subgroup migrating and the functional transformation is proposed. The main idea is, all peaks of multi-peak problems whose peaks are not equally high are transformed into those whose peaks are equally high by functional transformation so as to find all peaks in the same probability; Some genetic operators and near relative excluding strategy are executed in order to maintain the population diversity; The excellent individuals whose fitness are bigger than a threshold are migrated into a subpopulation; Gradient operator in subpopulation is applied to make the individuals evolved subtly. This searching strategy not only can ensure to find all peaks, but it doesnot need any pre-knowledge hypothesis, such as the number and distribution of the peaks. Finally, a comparison test with Spears' s Simple Sub Population strategy is performed.