摘要
针对PSO聚类算法需要预定聚类中心个数的问题,提出一种变维搜索解空间的量子粒子群优化聚类算法.该算法采用量子编码的方式实现双链并行搜索,加速寻优过程,避免了粒子在解空间边界过分聚集;设计了幅角相位旋转算子和变异算子,使幅角相位依变概率进行变异,提高了粒子群的多样性;在迭代过程中,动态更新了聚类中心的数量,使算法能够在不同维度的解空间中寻优.仿真实验表明,该算法的收敛速度和聚类精度得到一定的改善.
To solve the problem of predefining clusters amount in PSO, a novel Quantum PSO clustering algorithm based on variable di mensions search in solution space was proposed. The algorithm adopted quantum state encoding to implement hi-chain parallel searching, accelerating the speed of obtaining optimum and avoiding the over-congregating particles near boundaries of solution space; designed rotation operator and mutation operator, mutating particles in a variable probability, which enhanced the diversity of swarms effectively; updated the number of cluster centroids during the process of iteration dynamically, making the algorithm to search optimum in different dimensions. Simulation experiment indicates that the algorithm improves the converge speed and cluster accuracy to some extent.
出处
《小型微型计算机系统》
CSCD
北大核心
2012年第4期804-808,共5页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(60972161)资助
关键词
量子粒子群优化
动态聚类
双链并行搜索
变维搜索
quantum PSO
clustering dynamically
bi-chain parallel searching
var-dimension searching
