摘要
介绍了谱聚类方法的基本原理和算法思想,针对谱聚类方法优化问题求解的困难,分析了一种有原则的求解策略,从而给出算法的具体描述,并作为一个插件在Weka上进行了实现。对实现的系统进行了实验和测试,指出了应用中的关键问题。实验结果表明,谱聚类方法效果优于K-means方法。
The principle of spectral clustering and the thinking of algorithms are introduced, and a solving strategy is analyzed for dealing with the hard difficulty ofthe spectral clustering optimal problem. So, a detail description of spectral clustering algorithm is presented, and the algorithm is implemented in Weka as a plug-in component. Also, the key problems in application are pointed. The results of experiments show that spectral clustering has better performance than K-means clustering.
出处
《计算机工程与设计》
CSCD
北大核心
2008年第13期3384-3386,3421,共4页
Computer Engineering and Design
基金
广西自然科学基金项目(桂科自0481016)
广西教育厅2006年科研基金项目(149)
关键词
谱聚类
图分割
优化离散解
Weka接口
谱聚类类
spectral clustering
graph partition
optimal discretization solution
Weka interface
spectral clustering class