摘要
模糊C均值聚类对初始参数有着较强的依赖性,文中针对其对初始聚类中心敏感的问题,提出利用量子粒子群来优化FCM的初始聚类中心。粒子群优化算法具有较强的全局搜索能力,但局部搜索能力不足,因此借助于量子理论,将粒子群量子化,借助量子旋转门改变粒子的移动,同时利用量子非门增加种群的多样性,加强粒子群优化算法的局部寻优能力。并最终利用量子粒子群优化算法搜寻FCM算法的初始聚类中心,通过实验仿真表明,改进的算法在加快搜索速度的同时,能获得较为稳定的聚类中心且分割效果明显优于标准的FCM算法。
Fuzzy c-means （FCM） clustering has an excessive dependence on initial parameters. The quantum particle swarm is adopted to optimize initial clustering center of FCM to address its sensitivity to the initial clustering center. The particle swarm optimization algorithm has stronger global searching ability, but insufficient local search ability, The quantum rotating gate changes the movement of the particles and increases the diversity of population, thus better local optimization ability of particle swarm optimization algorithm. The quantum particle swarm optimization algorithm is used to search the initial clustering center of FCM algorithm. The experimental simulation shows that the improved algorithm can speed up the search while obtaining more stable clustering center and better segmentation effect than the standard FCM algorithm segmentation effect.
出处
《电子科技》
2016年第11期137-141,共5页
Electronic Science and Technology
