Multi-view clustering via consensus coefficient matrix and separate segmentation matrices

In recent years, achieving data from different sources and different views has caused to have many multi-view data sets. Among multi-view learning methods, multi-view clustering has been considered as an appropriate method to analyse these data by many researchers. The purpose of multi-view clustering is to use complementary information from all views to have clustering on multi-view data set. In this paper, we propose a spectral based multi-view subspace clustering method. Subspace clustering displays data in low dimensional spaces by using self-expressive properties, and finally the spectral clustering method is used for the final clustering of the data. In recent multi-view clustering methods, fusion matrix of segmentation or coefficient matrices are considered separately. In our proposed method, our aim is to obtain a consensus matrix of eigenvectors by joint learning via segmentation and coefficient matrices of all views, simultaneously. After that, spectral clustering over the consensus segmentation matrix would give a multi-view subspace clustering. At the end, we evaluate our method with recent methods those are related to our technique.