基于时变Sigmoid 函数的鲁棒PSO 算法

在样本规模有限的情况下, 为了提高算法的鲁棒优化性能, 提出一种基于时变(随迭代次数变化) Sigmoid 函数的鲁棒粒子群优化算法. 采用拟蒙特卡罗积分方法近似估计有效目标函数, 以时变Sigmoid 函数为基础, 设计各代各样本规模的选取概率. 迭代前期, 样本规模期望值较小, 加快了算法探索速度; 迭代后期, 样本规模期望值较大, 提高了算法的开发精度. 标准测试函数仿真结果显示, 所提出方法具有较优的鲁棒优化性能.

Robust particle swarm optimization algorithm based on time-varying Sigmoid function

To enhance the searching capability with the limited sample scale, a sort of robust particle swarm optimization algorithm based on time-varying Sigmoid function is proposed. Quasi-Monte Carlo method is used to approximate the effective objective function. The selecting probability of different sample size in different iteration is designed based on the time-varying Sigmoid function, which is changed in the iteration process. In the prophase of the algorithm, the expected value of sample size is small, and then the exploration speed is accelerated. In the anaphase of the algorithm, the expected value of sample size is large, and then the exploitation precision is improved. The simulation results of standard test functions show that this method possesses better robust optimization capability.