基于新搜索策略的混合蛙跳算法  

SHUFFLED FROG LEAPING ALGORITHM BASED ON NEW SEARCH STRATEGY

作　　者：赵芳[1] 张桂珠[1] 

机构地区：[1]江南大学物联网工程学院,江苏无锡214122

出　　处：《计算机应用与软件》2015年第8期224-228,共5页Computer Applications and Software

基　　金：国家自然科学基金项目(61170120);江苏省自然科学基金项目(BK2011147)

摘　　要：混合蛙跳算法(SFLA)具有算法简单、控制参数少、易于实现等优点,但在高维优化问题中算法易早熟收敛且求解精度低。为此,提出一种基于新搜索策略的混合蛙跳算法(NSSFLA)。该算法定义了新的粒子分类标准,将所有青蛙按此标准进行分类,每类青蛙按照相应的位置更新公式进行更新;在迭代过程中,每个青蛙个体根据自身状态动态地调整惯性权重,平衡了算法全局搜索和局部搜索的能力;在全局迭代中借鉴柯西变异优化策略思想,并以停滞代数判断是否对最优个体进行优化,避免了族群陷入局部最优。实验仿真表明,NSSFLA的寻优能力强,迭代次数少,解的精度高,更适合高维复杂函数的优化。Shuffled frog leaping algorithm（ SFLA） has the advantages of simple algorithm,less control parameters,and easy to realise,but in high-dimensional optimisation problem,it is easy to be premature convergence and the solution has low precision as well. Therefore,this paper proposes a new search strategy-based shuffled frog leaping algorithm（ NSSFLA）,the algorithm defines a new classification criterion of particles,all the frogs are classified according to the criterion,each species update according to the corresponding position updating formula. In the iterative process,each individual dynamically adjusts the inertia weight according to its own status,which balances the ability of algorithm in global search and local search. In global iteration,the thought of Cauchy mutation optimisation strategy is used as the experience,and whether or not to optimise the best individual is determined by stagnation algebraic,which avoids the population to fall into local optimum. Finally,simulation experiments show that NSSFLA has strong search capability,less number of iterations and higher precision in solution,and is more suitable for the optimisation of complex functions with high dimensionality.

关 键 词：群智能算法 混合蛙跳算法 柯西变异 分类标准 搜索策略 惯性权重 全局优化 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]