Extraction of Fuzzy Clusters from Weighted Graphs
Extraction of Fuzzy Clusters from Weighted Graphs
复制标题
从加权图中提取模糊簇
DOI：
10.1007/3-540-45571-x_51
复制
发表时间：
2000
期刊：
影响因子：
--
通讯作者：
K. Urahama
中科院分区：
文献类型：
--
作者：
S. Hotta;K. Inoue;K. Urahama
提出了一种基于加权邻接矩阵的谱图划分方法。从一个节点集合中提取一个模糊聚类是由一个特征值问题，聚类是从大类到小类依次提取的。首先设计了无向图的聚类方案，然后将其扩展到有向图和无向二部图。这些聚类方法被应用于Web网络中的链接结构的分析和通过关键字或样本图像查询的图像检索。聚类的提取结构是可视化的多元探索方法称为对应分析。
A spectral graph method is presented for partitioning of nodes in a graph into fuzzy clusters on the basis of weighted adjacency matrices. Extraction of a fuzzy cluster from a node set is formulated by an eigenvalue problem and clusters are extracted sequentially from major one to minor ones. A clustering scheme is devised at first for undirected graphs and it is next extended to directed graphs and also to undirected bipartite ones. These clustering methods are applied to analysis of a link structure in Web networks and image retrieval queried by keywords or sample images. Extracted structure of clusters is visualized by a multivariate exploration method called the correspondence analysis.