基于动态邻居和变异因子的多目标粒子群算法*

为了克服粒子群算法求解多目标问题极易收敛到伪Pareto前沿(等价于单目标优化问题中的局部最优解)和收敛速度较慢的缺陷，提出一种合并帕累托占优概念到动态邻居和变异因子的粒子群算法(particle swarm optimizer based on dynamic neighbor topology and mutation operator，DNMPSO)来处理多目标优化问题(DNMMOPSO),该算法也合并了外部存档技术来存储每次迭代产生的非劣解。模拟结果表明，提出的算法在多目标检测问题上要优于其他算法

Multi-objective particle swarm optimizer based on dynamic neighbor and mutation operator

In order to conquer multi-objective particle swarm optimizers(MOPSOs) easily converge to a false Pareto front (i.e.,the equivalent of a local optimum in single objective optimization), and converge slowly,this paper combined the Pareto dominance to DNMPSO to deal with multi-objective problems,and emploied the external archive to store the non-dominated solution at each iteration. Simulation results show that the proposed algorithm is able to find better solutions compared against other algorithms.Consequently, DNMMOPSO can be used as an effective algorithm to solve multi-objective problems.