Incomplete multiview subspace clustering algorithm based on self-representation and consistent learning

Incomplete multiview clustering is an important topic in machine learning due to ubiquitous incompleteness of view in the real data in life. The incompleteness of views is prevalent in real life. In this paper, we propose an incomplete multiview subspace learning method based on self-representation learning and graph fusion. The proposed method construct the robust graph of each view through the self-expression and semi-nonegative matrix decomposition, and learn a consistent representation of all views by Laplace co-regularization based on the low-dimensional representation of the each view. Through the learned consistent representation, the incomplete multiview can be divided by k-means. Experiments is implemented on 3 benchmark datasets and the superior results validate the effectiveness of the proposed method.