自适应多普勒补偿与变异选择的蝙蝠算法

为克服蝙蝠算法在高维优化问题上求解精度低和早熟收敛的缺点,提出一种改进的蝙蝠算法。首先根据蝙蝠相对猎物距离的远近程度,对频率引入自适应多普勒补偿策略,并结合速度偏移机制修正飞行方向,产生靠近最优个体的新位置;其次对最优个体构造自适应变异选择策略,先利用柯西变异产生的较大步长摆脱局部极值的束缚,后利用高斯变异产生的较小步长精细搜寻最优区域;最后通过调整响度和脉冲发射率,平衡算法的全局探索和局部开发能力。从理论上分析了算法的收敛性和运算复杂性,对12个标准函数在不同维度下进行仿真实验,并与近年来其他蝙蝠算法进行比较,结果表明改进的算法在求解高维优化问题上具有较优的收敛速度和精度。

Self-Adaptive Doppler Compensation and Mutation Choice of Bat Algorithm

This paper presents an improved bat algorithm(Doppler and mutant bat algorithm, DMBA) to overcome the problems of high dimensional optimization with low precision and premature convergence. A self-adaptive compensation for the Doppler effect of frequency is introduced according to the relative distance between the bat and the prey, and the flight direction is modified by combining the velocity offset mechanism to generate a new position close to the prey. Then, self-adaptive mutation choice strategies are constructed for the best individual. The larger step size generated by Cauchy mutation is used to get rid of the constraint of local extreme value, and then the smaller step size generated by Gaussian mutation is used to search the optimal region. Finally, the global exploration and local exploitation ability are balanced by adjusting loudness and pulse emission. The convergence and computational complexity of the algorithm are analyzed theoretically. 12 classical benchmark functions are simulated in different dimensions and the proposed algorithm is compared with other recent bat algorithms. The experimental results show that the proposed algorithm has better convergence speed and precision for solving high dimensional optimization problems.