基于特征方程的蝙蝠算法分析及其改进策略

Characteristic equation-based and its analysis of bat algorithm improving strategies

在求解复杂非线性优化问题时,蝙蝠算法因其进化机制中引入了更多可调参数因子而比粒子群算法和遗传算法等具有更好的收敛性能.然而,在其迭代过程中,一旦群体中出现"超级"蝙蝠个体,算法极易出现"迟滞"问题.针对该问题,采用特征方程方法对基本蝙蝠算法的收敛性进行了分析,在一定假设条件下,讨论了算法参数灵敏性.基于负梯度理论,通过调整算法中蝙蝠个体的速度更新策略,使其沿群体当前最优解的负梯度方向飞行,引导个体飞向全局最优解.典型benchmark函数仿真实验结果表明,改进蝙蝠算法表现出较基本蝙蝠算法和带速度权重的改进粒子群算法更好的全局寻优能力.

Bat algorithm always outperforms particle swarm optimization and genetic algorithm in solving complex nonlinear optimization problems for there are moreadjustable parameters controlling its evolutionary rules.However,bat algorithm tends tobe troubled by ＂premature＂once a ＂superior＂individual bat from the bat population is trapped into a local optimal solution.To improve the performance of the original bat algorithm,the convergence is analyzed by characteristic equation method,and the sensibility of the parameters is discussed under a postulated condition firstly.An improved bat algorithm is then presented by regulating the velocity updating strategy of individual bat based on negative gradient theory to lead the batto＇fly＇to the global solution towardsthe negative gradient direction of the current optimal solution.The experimental results based on typical benchmark functions show that the improved bat algorithm achieves better optimization solution than the original bat algorithm and weighted particle swarm optimization.