《计算机应用研究》|Application Research of Computers

标记判别和局部线性强化的半监督稀疏子空间聚类

Semi-supervised sparse subspace clustering based on label discrimination and local linear reinforcement

免费全文下载 (已被下载 次)  
获取PDF全文
作者 朱恒东,马盈仓
机构 西安工程大学 理学院,西安 710600
统计 摘要被查看 次,已被下载
文章编号 1001-3695(2021)10-022-3014-05
DOI 10.19734/j.issn.1001-3695.2021.03.0044
摘要 子空间聚类通常可以很好地处理高维数据,但由于数据本身的噪声等的影响,系数矩阵的块对角线结构往往容易被破坏。针对上述问题,提出了一种标记判别和局部线性强化的半监督稀疏子空间聚类。一方面,通过约束标记数据之间的系数为0,更好地捕获数据的全局结构;另一方面,通过K近邻关系加强数据邻近点之间的局部相关性,同时消除大量不相关的数据点,增强算法的鲁棒性。通过在多种数据上的实验,验证了提出的半监督聚类算法的有效性。
关键词 子空间聚类; K近邻; 半监督; 稀疏
基金项目 国家自然科学基金资助项目(61976130)
陕西省重点研发计划资助项目(2018KW-021)
陕西省自然科学基金资助项目(2020JQ-923)
本文URL http://www.arocmag.com/article/01-2021-10-022.html
英文标题 Semi-supervised sparse subspace clustering based on label discrimination and local linear reinforcement
作者英文名 Zhu Hengdong, Ma Yingcang
机构英文名 School of Science,Xi'an Polytechnic University,Xi'an 710600,China
英文摘要 Subspace clustering can usually handle high-dimensional data well, but due to the influence of the noise of the data itself, the block diagonal structure of the coefficient matrix is often easily destroyed. To solve the above problems, this paper proposed a semi-supervised sparse subspace clustering with label discrimination and local linear reinforcement. On the one hand, it better captured the global structure of the data by constraining the coefficient between the labeled data to be 0. On the other hand, it strengthened the local correlation between the neighboring points of the data through the K-nearest neighbor relationship, and eliminated a large number of unrelated data points enhance the robustness of the algorithm. The experiments on a variety of data verify the effectiveness of the proposed semi-supervised clustering algorithm.
英文关键词 subspace clustering; K-nearest neighbor; semi-supervised; sparse
参考文献 查看稿件参考文献
 
收稿日期 2021/3/17
修回日期 2021/4/25
页码 3014-3018,3034
中图分类号 TP181
文献标志码 A