基于云推理的协方差矩阵自适应进化策略算法

IMPROVED CMA-ES ALGORITHM BASED ON CLOUD REASONING

针对协方差矩阵自适应进化策略(CMA-ES)在求解某些问题时存在早熟收敛、精度不高等缺点,通过利用云模型良好的不确定性问题处理能力对CMA-ES的步长控制过程进行改进,得到一种基于云推理的改进CMA-ES算法。该算法通过建立步长控制的云推理模型,采用云模型的不确定性推理来实现步长的控制,避免了原算法采用确定的函数映射进行步长伸缩变化而忽视进化过程中不确定性的不足。最后通过测试函数验证了改进算法具有较高的寻优性能。

In order to overcome the shortcom ings o f covariance m a trix adaptation evo lu tion strategy (C M A -E S ) such as prem ature convergenceand low precision when being used in some o p tim isation p ro b le m s ,b y m a kin g use o f the good a b ility o f cloud m odel in dealing w ith u n certa in ty p ro b le m s,w e im prove the step-size control process o f C M A -E S and fin d a cloud reasoning-based im proved C M A -E S a lg o rith m . A fte rb u ild in g the cloud reasoning m odel o f step-size c o n tro l, the im proved a lgo rithm achieves step-size con trol by using u n ce rta in ty reasoning ofcloud m o d e l,a n d avoids the d e ficie n cy o f o rig in a l a lgo rithm that it uses d e te rm in istic fu n c tio n m apping fo r step-size scale b u t ignores the u n certa in ty in evo lu tio n process. F in a lly , through test fu n ctio n s we v e rify th a t the im proved a lgo rithm has h ig h e r o p tim isation perform ance.