Multi-View Subspace Clustering With Block Diagonal Representation

Jipeng Guo,Wenbin Yin,Yongli Hu,Yanfeng Sun

doi:10.1109/access.2019.2923614

Abstract

Self-representation model has made good progress for a single view subspace clustering. This paper proposed the multi-view subspace clustering model based on self-representation. This model assumes that the samples from different classes are embedded in independent subspaces. Thus, the fused multi-view self-representation feature should be block diagonal, and a block diagonal regularizer with the complementarity of multi-view information is given. The model optimization algorithm by alternating minimization is proposed and its convergence without any additional assumption is proved. With the complementarity of multi-view information and the block diagonal property, our model will depict data more comprehensively than single view independently. The extensive experiments on public datasets demonstrate the effectiveness of our proposed model.

Highlights

Data clustering divides similar samples together by exploring the intrinsic structure between data, which is an important and fundamental topic of unsupervised learning
Motivated by single-view representations model, this paper proposed a block diagonal representation model for multi-view subspace clustering named as MSCBDR
We compare our method with following baselines: co-regularized spectral clustering(CRSC) [10], multi-modal spectral clustering(MMSC) [12], parameter-weighted multiview clustering(PwMC) [26], self-weighted multi-view clustering (SwMC) [26] and latent multi-view subspace clustering(LMSC) [33]

Summary

INTRODUCTION

Data clustering divides similar samples together by exploring the intrinsic structure between data, which is an important and fundamental topic of unsupervised learning. The Hadamard product on all single graphs is employed to obtain precise internal structure These multi-view clustering algorithms can effectively mine multiple heterogeneous information and get better performance. Guo et al.: Multi-View Subspace Clustering With Block Diagonal Representation view-collaborative, attribute-weighted MEC (VC-AW-MEC) are proposed in [52] Another multi-view clustering algorithm is based on Multiple Kernel Learning [22], [23], which integrates different views by composing the different kernels and is suitable for clustering large scale and high dimension data without constructing complex similarity graph. Motivated by single-view representations model, this paper proposed a block diagonal representation model for multi-view subspace clustering named as MSCBDR. A natural multi-view fusion method is proposed, and block-diagonal constraints on multi-view representation matrices are used to obtain accurate heterogeneous information.

BLOCK DIAGONAL REPRESENTATION FOR MULTI-VIEW

CLUSTERING TASKS

COMPUTATION COMPLEXITY

EXPERIMENTS

EXPERIMENTS SETTING

CONCLUSIONS