Deep manifold regularized semi-nonnegative matrix factorization for Multi-view Clustering

Xiangnan Liu, Shifei Ding, Xiao Xu, Lijuan Wang

Abstract: Highlights•We use both the similarity and exclusion between objects to construct the manifold regularizer, which avoids the excessive dependence of the geometric structure information on the nearest neighbors.•We consider the multi-layer manifold regularization to ensure the stability of the internal geometric structure and strengthen the accuracy of the new representation matrix generated by the Deep Semi-NMF model.•We add auto-weights to distinguish the value of information in each view and highlight the difference and complementarity of multi-view data.