基于多样性的多视图低秩稀疏子空间聚类算法

Multiview low-rank sparse subspace clustering based on diversity

本文主要研究如何通过挖掘多视图特征的多样性信息来促进多视图聚类,提出了基于多样性的多视图低秩稀疏子空间聚类算法。该方法直接将视图多样性概念应用于多视图低秩稀疏子空间聚类算法框架中,确保不同视图的子空间表示矩阵的多样性;为了实现多个视图聚类一致性同时达到提高聚类性能的目标,在该框架中引入谱聚类算法共同优化求解。通过对3个图像数据集的实验验证了该算法的有效性,同时其聚类的性能优于已有的单视图及多视图算法。

This paper focuses on boosting multiview clustering by exploring the diversity of information among multiview features.A multiview clustering framework,called multiview low-rank sparse subspace clustering based on diversity,is proposed for this task.In the proposed method,the concept of diversity is successfully introduced into the framework of a multiview low-rank sparse subspace clustering algorithm to ensure that the representation matrix of different subspace views has certain differences and that the obtained information is diversified.In addition,the spectral clustering algorithm is added to the framework to achieve a joint optimization solution,which can markedly improve the clustering performance,to obtain a unified target clustering assignment.The effectiveness of the algorithm is verified by fully observing three image datasets,and the clustering performance of the proposed algorithm is better than that of the existing single-view and multiview algorithms.