自适应变异的天牛群优化算法

Beetle swarm optimization algorithm with adaptive mutation

针对复杂优化问题下粒子群优化算法收敛速度慢和易早熟收敛的缺陷,提出一种结合粒子群优化算法与天牛须搜索算法的新型优化算法——自适应变异的天牛群优化算法。首先,通过引入个体粒子对周围环境的感知机制,构造一种基于自适应须长与步长的天牛群优化算法,丰富个体在迭代过程中可参考的信息;然后,引入多维扰动群体最优位置的变异策略,实现减少陷入局部最优解的功能;最后,根据群体聚集程度调整变异概率,并随着迭代的进行逐步降低变异概率以使天牛群在迭代后期稳定在局部精细搜索。为验证算法的性能,将新算法与其他7个对比算法针对7个经典测试函数在不同维度下针对平均适应值和算法运行时间进行比较,此外,还进行了神经网络训练对比测试以验证算法实用效果。实验结果表明新算法的寻优效果和收敛速度较其他算法有较大提高,尤其适合应对高维复杂优化问题。

Considering the defects of slow convergence and easy premature convergence of particle swarm optimization algorithm for complex optimization problems,a beetle swarm optimization algorithm with adaptive mutation was proposed based on ideas of particle swarm optimization algorithm and beetle antennae search algorithm.Firstly,by introducing a perception mechanism of individuals to surrounding environment,a beetle swarm optimization algorithm based on adaptive length of antennae and step size was constructed to enrich information the individual referenced in iterations.Secondly,based on multi-dimensional disturbance to the best position of group,a mutation strategy was used to avoid falling into local optimum.Finally,the mutation probability was adjusted according to group aggregation,and gradually reduced with the iteration to enhance the local search ability of beetle swarm late in the iteration.To verify the performance,the new algorithm was compared with the other seven comparison algorithms on seven classical test functions with different dimensions,in average fitness and algorithm runtime.In addition,a comparison test of neural network training was carried out to verify the practical effect of the algorithm.The experimental results show that the new algorithm has better optimization effect and higher convergence speed than those of other algorithms,especially for complex and high-dimensional optimization problems.