一种基于精英选择和反向学习的分布估计算法

An Improved Estimation of Distribution Algorithm with Extreme Elitism Selection and Opposition-Based Learning

为了提高分布估计算法的性能,提出一种从选择方式和搜索机制出发的改进分布估计算法。首先,借鉴并改进粒子群种群更新策略,在构造优势群体时,引入精英选择策略替换经典的截断选择,提高算法向全局最优解的收敛速度;然后,引入二次反向反射搜索算子,从搜索机制上对分布估计算法进行改进,提高算法的全局搜索能力。仿真结果表明,所提出的改进分布估计算法能够有效的避免陷入局部最优值,在寻优精度、收敛速度以及算法的稳定性和鲁棒性上都有极大改善。

An existing estimation of distribution algorithm(EDA)with univariate marginal Gaussian model was improved from the perspective of selection mode and search mechanism in this study.Frist,an extreme elitism selection method was incorporated into the EDA to obtain a faster convergence rate.The selection method was inspired by the population regeneration mechanism of the particle swarm optimization(PSO)algorithm.It accelerated convergence to the optimal solution by highlighting the role of a few top best individuals when constructing the selected population which is the basis of the probabilistic model.Furthermore,a quasi-reflection opposition-based learning was introduced to improve the global search ability of the EDA.Simulations show that the proposed EDA with extreme elitism selection and quasi-reflection opposition-based learning(EEQO-EDA)significantly outperforms EDA with extreme elitism selection(EE-EDA)which has been reported in the literature and the standard EDA in terms of the effectiveness and quality of the solution.