摘要
半监督聚类利用少部分标签的数据辅助大量未标签的数据进行非监督的学习,从而提高聚类的性能。提出一种基于谱聚类的半监督聚类算法,其利用标签数据的信息,调整点与点之间的距离所形成的距离矩阵,而后基于被调整的距离矩阵进行谱聚类。实验表明,该算法较之于已提出的半监督聚类算法,获得了更好的聚类性能。
Semi-supervised clustering employs a small amount of labeled data to aid unsupervised learning. In this paper a new semi-supervised clustering method based on spectral clustering was proposed. Making use of the information the labeled data contains, the distance matrix derived from data was modified and then the spectral clustering method was uesed to get the final clusters according to the modified distance matrix. Experimental result demonstrates that compared with previously proposed semi-supervised clustering algorithm this method produces better clusters.
出处
《计算机应用》
CSCD
北大核心
2005年第6期1347-1349,共3页
journal of Computer Applications
关键词
半监督聚类
谱聚类
semi-supervised clustering
spectral clustering