Multi-view clustering with interference based on cross-view mapping

Multi-view clustering (MVC) has attracted widely attention, and traditional methods often suffer from the challenging problem of data interference. To deal with this issue, a novel MVC method based on cross-view mapping scheme is proposed in this paper. In our method, the data interence problem is studied using neighborhood information and the cross-view mapping matrix is designed to recover the interfered data. An efficient model is constructed, where the geometric structure of data is exploited. Finally, an alternative iteration updating algorithm is developed to solve this model, together with the study of the corresponding computational complexity and the convergence. Experimental results on different datasets demonstrate that our method outperforms the compared methods.