优化算法中均值信息利用研究

Research on the Utilization of Mean Value Information in Optimization Algorithm

研究了启发式优化算法中种群向均值点迁移的策略,并发现该策略对于提升算法性能具有重要影响,同时具备物理和数学含义.通过极大似然估计方法对基态波函数进行参数估计,建立了量子系统达到基态时最优解概率密度函数与种群均值点之间的联系,并从动力学的角度解释了种群均值点的物理意义.通过在几种经典优化算法上添加利用均值点位置信息的操作,在CEC2013测试集与摄像机布局优化的工程应用上进行对照实验,实验结果表明合理利用均值点位置信息可以有效提升算法的性能.

The strategy of population migration towards the mean point in heuristic optimization algorithms is investigated,and the strategy is found that it has significant impact on the algorithm’s performance and has both physical and mathematical implications.By using the maximum likelihood estimation method,the parameters of the ground state wave function are estimated,and the connection between the probability density function of the optimal solution when the quantum system reaches the ground state and the population mean point is established.The physical significance of the population mean point is explained from a dynamic perspective.The operations that utilize the information of the mean point’s position is added to several classical optimization algorithms and a comparative experiment is carried out on the CEC2013 test set and the engineering application of camera layout optimization.Experimental results show that reasonable use of the mean point position information can effectively improve the performance of the algorithm.