摘要： 根据聚类假设，提出一种新的基于图的半监督学习算法，称为密度敏感的半监督聚类。该算法引入一种密度敏感的距离测度，它能较好地反映聚类假设，并且充分挖掘了数据集中复杂的内在结构信息，同时与基于图的半监督学习方法相结合，使得算法在聚类性能上有了显著的提高。经过实验仿真进一步表明，该算法在特定图像应用上具有优越性。

关键词: 半监督聚类, 密度敏感, 聚类假设

Abstract: Based on clustering assumption, this paper proposes a novel semi-supervised learning algorithm based on graph, named Density- sensitive Semi-supervised Clustering(DS-SC). The approach introduces a density-sensitive distance measure which reflects the clustering assumption well and fully exploits the competitive inherent structure information among dataset, and combining it with a graph-based semi-supervised learning methods leads to prominent clustering performance enhance of DS-SC. The results demonstrate the superiority of DS-SC in the application of a specific image in the further simulation.

Key words: semi-supervised clustering, density-sensitive, clustering assumption

中图分类号: 

• TP311