摘要
为了获得结构更加合理的相似矩阵,提出了基于谱聚类和L_(2.1)范数的多视图聚类算法。该算法首先将改进的多视图亲和矩阵利用L_(2.1)范数正则项合理地构造出相似矩阵S,使S在整体稀疏的情况下保证局部的强线性关系;然后同时进行相似矩阵的学习和谱聚类过程,将相似矩阵S和标签矩阵F交替迭代,加强数据集与降维后的F的几何结构的紧密联系;最后对所提出的算法进行了实验,结果表明该算法是有效的。
In order to obtain a more reasonable structure of similarity matrix,a multi-view clustering algorithm based on spectral clustering and L_(2.1)norm is proposed.The algorithm first uses the improved multi-view affinity matrix to reasonably construct a similarity matrix S using the regular term of the L_(2.1)norm,so that S is sparse as a whole to ensure a strong local linear relationship,and then simultaneously performs similarity matrix learning and spectral clustering in the process,the similarity matrix S and the label matrix F are alternately iterated to strengthen the close connection between the data set and the geometric structure of F after dimensionality reduction.Finally,experiments are carried out on the proposed algorithm,and the results show that the algorithm is effective.
作者
贺娜
马盈仓
张丹
续秋霞
HE Na;MA Yingcang;ZHANG Dan;XU Qiuxia(School of Science,Xi'an Polytechnic University,Xi'an 710600)
出处
《计算机与数字工程》
2021年第11期2335-2341,共7页
Computer & Digital Engineering
基金
国家自然科学基金项目(编号:61976130)
陕西省重点研发计划项目(编号:2018KW-021)
陕西省自然科学基金项目(编号:2020JQ-923)资助。
