基于中值迭代函数的自适应序列生境粒子群优化算法

A NOVEL ADAPTIVE SEQUENTIAL NICHE TECHNIQUE FOR PARTICLE SWARM OPTIMIZATION ALGORITHM BASED ON RECURSIVE MIDDLING FUNCTION

提出一种新的基于中值迭代函数的自适应序列生境粒子群优化算法。该算法利用中值迭代函数来判断搜索空间中的任意两点是否属于相同的峰,从而自适应地改变当前进化粒子的适应值,克服了标准序列生境算法中必须利用先验知识确定小生境半径的缺陷以及在利用山谷函数分类中必须利用先验知识确定采样概率矩阵的缺陷。将该算法用于多峰函数最优搜索问题。通过多个Matlab仿真实验,验证了算法的有效性。实验结果表明:算法能够自适应、更高效准确地遍历多峰函数的所有极值,可应用于求解局部最优和全局最优问题。

In this paper it proposes a novel adaptive sequential niche particle swarm optimization algorithm, which is based on the Recursive Middling （ RM ） function. In this algorithm, the RM function is used to determine whether any two points in search space belong to the same peak of the multimodal function or not, and then to change adaptively the fitness of a current evolutionary particle in a sub-swarm. It overcomes the pitfall in standard sequential niche algorithm that niche radius have to be determined with prior knowledge and the pitfall in the use of hill-valley function that the sampling probability array has to be decided with prior knowledge. Applying this algorithm to searching the multiple optimal solutions for multimadal function,its validity has been proved by a couple of Matlab emulation experiments. The comparative experimental results show that the proposed algorithm is able to traverse all the extrema of multimodal function adaptively and efficiently, and to be applied in searching local and global multiple optimal solutions for the benchmark test function.