嵌入自学习策略的多因子进化算法

Multifactorial evolutionary algorithm with embedded self-learning strategy

为了解决多因子进化算法(MFEA)局部搜索能力不足的缺陷,提出将基于拟牛顿方法的自学习策略嵌入到MFEA中.结合MFEA的特性,设计了三种类型的嵌入策略,并应用于多任务优化问题的求解.结果表明:嵌入式MFEA的性能远优于同样基于拟牛顿方法改进的自适应模因多因子进化算法(AMA-MFEA);在相同迭代次数的条件下,完全子代嵌入可以发挥最好的效果;在相同优化时间的约束下,随机抽样嵌入最为有效.

In order to solve the shortcomings of multifactorial evolutionary algorithm(MFEA) of insufficient local search ability,a self-learning strategy based on the quasi-Newton method was proposed to be embedded in the MFEA.Combining the characteristics of MFEA, three types of embedding strategies were designed and applied to the solution of multi-task optimization problems.Experimental results show that the performance of embedded MFEA is far superior to the adaptive memetic algorithm-multifactorial evolution algorithm(AMA-MFEA), which is also improved based on the quasi-Newton method. Research shows that under the condition of the same number of iterations, subpopulation embedding can exert the best effect;under the constraints of the same optimization time,random sampling embedding is the most effective.