吉林大学学报（理学版）2025，Vol.63Issue(2)：499-512,14.DOI:10.13413/j.cnki.jdxblxb.2024164
基于多样性和谱嵌入的张量多视图子空间聚类
Tensor Multi-view Subspace Clustering Based on Diversity and Spectral Embedding
摘要
Abstract
Aiming at how to effectively utilize the diversity information and higher-order information of multi-views,and establish the connection between the learning process of coefficient matrices and spectral clustering,we proposed a tensor multi-view subspace clustering algorithm based on diversity and spectral embedding.Firstly,the algorithm used tensor adaptive log-determinant regularization in the self-representation tensor part,which could adaptively select the approximation function according to the size of the singular values.Secondly,the Hilbert-Schmidt independence criterion was used to measure diversity to ensure that the coefficient representation matrices from different views exhibited sufficient diversity.Thirdly,in order to avoid the spectral clustering process to be performed independently,it was introduced into the model for joint learning,so that the low-rank tensor learning,diversity learning and spectral embedding learning were performed in a unified framework.Finally,the effectiveness of the algorithm in improving the clustering performance was verified by comparing it with ten excellent algorithms on five real datasets.关键词
多视图子空间聚类/张量自适应对数行列式/多样性/谱嵌入/Hilbert-Schmidt独立准则Key words
multi-view subspace clustering/tensor adaptive log-determinant/diversity/spectral embedding/Hilbert-Schmidt independence criterion分类
信息技术与安全科学引用本文复制引用
张沙沙,王长鹏..基于多样性和谱嵌入的张量多视图子空间聚类[J].吉林大学学报（理学版）,2025,63(2):499-512,14.基金项目
国家自然科学基金(批准号:12471480)和长安大学中央高校基本科研业务费专项基金(批准号:300102122101). （批准号:12471480）
