• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

计算机工程与科学

• 人工智能与数据挖掘 • 上一篇    下一篇

多路径高斯核模糊C均值聚类算法

文传军1,汪庆淼2   

  1. (1.常州工学院数理与化工学院,江苏 常州 213002;2.苏州大学计算机学院,江苏 苏州 215021)
  • 收稿日期:2016-08-22 修回日期:2017-01-03 出版日期:2018-05-25 发布日期:2018-05-25
  • 基金资助:

    国家自然科学基金(61170126);常州工学院校级课题(YN1305)

A multi-path Gaussian kernel
fuzzy C means clustering algorithm

WEN Chuan-jun1,WANG Qing-miao2   

  1. (1.School of Mathematical Sciences and Chemical Engineering,Changzhou Institute of Technology,Changzhou 213002;
    2.School of Computer Science and Technology,Soochow University,Suzhou 215021,China)
     
  • Received:2016-08-22 Revised:2017-01-03 Online:2018-05-25 Published:2018-05-25

摘要:

聚类算法单一迭代路径限制了参数优值的搜索。提出一种多路径高斯核模糊C均值聚类算法(MGKFCMs),MGKFCMs算法首先取核目标函数及模糊隶属度函数中的核函数为高斯核函数;然后利用梯度法得到聚类中心迭代公式,并基于该迭代公式和粒子群算法作聚类中心的并行参数迭代,在每一次聚类迭代时,选择聚类目标函数值小的路径作为参数迭代最终路径。对比分析了MGKFCMs算法的相关性质,通过仿真实验验证了所提算法的有效性。

关键词: 核方法, 模糊聚类, 高斯核, 聚类中心, 多路径迭代

Abstract:

The single iteration path of the clustering algorithm limits the search path of the parameter’s optimal value. In this paper, a multi-path Gaussian kernel fuzzy c-means clustering method is proposed and named MGKFCMs. Firstly, MGKFCMs takes the nuclear objective function and the fuzzy membership degree function in the kernel function as the Gaussian kernel function. Secondly, the gradient method is used to get the iterative formula of the clustering center. Based on this iterative formula and particle swarm optimization algorithm, the parameters of the clustering center are calculated iteratively in parallel. In every iteration of clustering, a path with small clustering objective function value is selected as the final path of parameter iteration. The correlation property of MGKFCMs is analyzed, and the convergence of the algorithm is studied. Simulation results show that the proposed algorithm is effective.
 
 

Key words: kernel method, fuzzy clustering, Gauss kernel, clustering center, multi-route iteration