组合分布估计和差分进化的多目标优化算法

Multi-objective optimization algorithm composed of estimation of distribution and differential evolution

为了提高多目标优化算法的收敛能力及求解精度,提出了一种组合分布估计和差分进化的多目标优化算法.该方法用分布估计算法和差分进化算法共同生成种群中的粒子,利用选择因子来控制每个粒子的产生方式,并且根据迭代次数的增加来改变2种算法的使用比例,搜索初期利用分布估计算法进行快速定位,然后用差分进化算法进行精确搜索.并对差分进化算法的变异因子进行了改进,定义了一个可变的变异因子,来控制不同搜索时期中差分进化算法的变异范围.用4个测试函数对算法进行了仿真测试,并同NSGA-Ⅱ和RM-MEDA进行了比较.实验结果表明,该算法具有良好的收敛性和分布性,并且效果稳定.

In order to improve the ability of convergence and accuracy of a multi-objective optimization algorithm, a multi-objective optimization algorithm composed of estimation of distribution and differential evolution has been pro-posed. Both estimation of distribution algorithm and differential evolution algorithm will be used to generate parti- cles of population. The generation method of each particle has been decided by using a selective factor, and propor- tion of the use of two algorithms according to the frequency of iterations. Utilizing an estimation of distribution algo- rithm to quickly locate in the initial search, and then differential evolution algorithm was used for accurately con- ducting searches. The variation factor of differential evolution algorithm was improved, and a variable variation fac- tor also was defined and used to control the range of variation of differential evolution algorithm in different search periods. Four test functions were used to evaluate the performance of the proposed algorithm, and next the proposed algorithm was compared with nondominated sorting genetic algorithm II （NSGA-II） and regularity model-based mul- tiobjective estimation of distribution algorithm （RM-MEDA）. The experimental results show that the proposed algo- rithm displayed a good convergence, diversity performance, and the stable effects.