一类约束不可微优化问题的极大熵萤火虫算法  

Maximum entropy firefly algorithm for a class of constrained non-differentiable optimization problem

作　　者：张建科 王高峰 尹露洋 

机构地区：[1]西安邮电大学理学院,陕西西安710121

出　　处：《西安邮电大学学报》2017年第5期94-100,共7页Journal of Xi’an University of Posts and Telecommunications

基　　金：陕西省教育厅科学研究计划资助项目(2013JK1130)

摘　　要：针对一类约束不可微优化问题,给出一种新混合算法。利用极大熵函数分别将非光滑目标函数和非光滑约束函数转化为光滑约束函数,构造目标函数与约束函数的增广拉格朗日函数,作为萤火虫算法的适应值函数加以优化,将其最优解近似代替原约束不可微优化问题的解,并对计算过程中参数值选取过大导致数据溢出的问题给出等效替换方案。实验结果显示,所给算法收敛速度快,数值精度高。In order to solve a class of constrained non-differentiable optimization problem,a new hybrid algorithm is proposed by combining the firefly algorithm and the maximum entropy function method.Firstly,the maximum entropy function method is used to transform the nonsmooth objective function and the non-smooth constraint function to a smooth unconstrained function,then the augmented Lagrange function of objective function and constraint function is constructed;Secondly,the augmented Lagrange function is used as the fitness function of the firefly algorithm,and the firefly algorithm is then used to optimize the fitness function;Finally,the optimum solution of this fitness function is used to replacement the solution of that constrained non-differentiable optimization problem.In the process of calculating,the parameter value is too large and this can because data overflow.An equivalent replacement scheme is therefore given to overcome this issue.Numerical results show that the new algorithm have high calculation precision and good numerical stability.

关 键 词：萤火虫算法 极大熵函数 约束不可微优化问题 数据溢出 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TP301.6[自动化与计算机技术—控制科学与工程]