改进粒子群算法在高维复杂函数寻优中的应用

针对粒子群算法应用于复杂函数优化时可能出现过早收敛于局部最优解的情况，提出了一种改进的算法结构，命名为多阶段多子群粒子群算法（Multi-stage Multi-subpopulation Particle Swarm Optimization，MMPSO），该方法主要通过多子群之间阶段性的重分组策略，强化不同群体之间的信息交流，增大其搜索到全局最优解的概率，同时，为了保留粒子群算法高效优化的特点，通过分阶段搜索模式的转变，将全局最好模型收敛的快速性和局部最好模型收敛的全局最优性进行折中，确保改进后的粒子群算法拥有更强的全局搜索能力和尽量高的收敛速度。仿真实验证明，MMPSO算法相对于基本粒子群算法而言具有明显的精度优势。

Improved particle swarm optimization algorithm for high-dimension complex functions

The PSO may convergence in local optimal solution when it is applied to optimization of complex functions,an improved structure of the algorithm,called multi-stage multi-subgroup PSO（Multi-stage Multi-subpopulation Particle Swarm Optimization,MMPSO） is constructed.It regroups the multiple subpopulations,achieves the information exchange of different groups,raises the efficiency of getting the global optimal solution.At the same time,in order to retain the efficiency of the algorithms,the transformation of the staged search model is made.It combines the rapidity of the global best model with the global optimality of the partial best model.These strategies ensure that the improved PSO has a stronger efficiency of getting the global optimal solution and the highest rapidity of convergence as far as possible.Then the test on the multiple peak value function demonstrates MMPSO algorithm has obvious advantage of accuracy.