基于密度自适应邻域相似图的半监督谱聚类.PDFVIP

37 9 Vol. 37 No. 9 第 卷第 期 计算机应用研究 录用定稿 Application Research of Computers Accepted Paper 基于密度自适应邻域相似图的半监督谱聚类 * 刘友超，张曦煌 (江南大学 物联网工程学院, 江苏 无锡 214122) 摘 要：谱聚类是基于谱图划分理论的一种聚类算法。传统的谱聚类算法属于无监督学习算法，只能利用单一数据 来进行聚类。针对这种情况，提出一种基于密度自适应邻域相似图的半监督谱聚类(DAN-SSC)算法。半监督学习是 介于监督学习和无监督学习之间的一种机器学习方法，主要研究的是如何同时且有效地利用无标签数据和有标签数 据来共同改进学习算法的性能。DAN-SSC 算法在传统谱聚类算法的基础上结合了半监督学习的思想，很好地解决 了传统谱聚类算法无法充分利用所有数据，不得不对一些有标签数据进行舍弃的问题；并且将少量的成对约束先验 信息扩散至整个空间，使其能更好地对聚类过程进行指导。实验结果表明，DAN-SSC 算法具有可行性和有效性。 关键词：谱聚类；密度自适应邻域；相似图；半监督学习 中图分类号：TP doi: 10.19734/j.issn.1001-3695.2019.04.0113 Semi-supervised spectral clustering based on density adaptive neighbor similarity graph Liu Youchao, Zhang Xihuang (School of Internet of Things Engineering , Jiangnan University, Wuxi Jiangsu 214122, China) Abstract: The spectral clustering is a clustering algorithm based on the theory of spectral partitioning . The traditional spectral clustering algorithm belongs to unsupervised learning algorithms and can only utilize a single type of data to cluster. Based on the situation, this paper proposed a semi-supervised spectral clustering algorithm based on density adaptive neighbor similarity graph(DAN-SSC) . Semi-supervised learning is a machine learning method between supervised learning and unsupervised learning. It focuses on how to use unlabeled data and labeled data simultaneously and effectively to improve the performance of learning algorithms. DAN-SSC algorithm combines the idea of semi-supervised l