结合梯度法的混合微粒群优化算法

在微粒群优化算法PSO中引入梯度算法,提出了一种新型的混合微粒群优化算法———GPSO。该混合优化算法是对PSO每一次进化后的所有微粒进一步执行梯度法寻优操作,并以寻找到的更优个体替代当前个体参与群体的下一代进化。GPSO既利用了PSO出色的全局搜索能力,又借助梯度法的快速局部寻优能力,很好地将两者的优势结合在一起。数值实验表明:无论是对于低维的多峰函数,还是高维的多峰和单峰病态函数,GPSO都表现出很强的优化效率、适用性和鲁棒性。

Hybrid particle swarm optimization algorithm based on gradient method

By introducing the gradient method into the particle swarm optimization algorithm (PSO),a new hybrid particle swarm optimization algorithm named GPSO is proposed.In GPSO,the gradient method is applied to each particle to search further for a better position after each evolution of PSO,and then the particle swarm replaced by the better takes part in the evolution of the next generation.The presented hybrid method incorporates the advantages of the excellent global searching of the PSO and the local speedy convergence of the gradient method.Several numerical examples are used to demonstrate the high optimization efficiency and the robustness of the given method for either low-dimensional multimodal functions or high-dimensional multimodal or pathological functions.