Tri-factor Nonnegative Matrix Factorization (TNMF) is of use in simultaneously clustering rows and columns of the input data matrix. In this paper, we present a Sparse Manifold Embedding Tri-factor Nonnegative Matrix Factorization (STNMF) for data clustering. Similar to most graph regularized NMF, STNMF is to extend the original TNMF by incorporating the graph regularized and sparse manifold embedding constraints into the TNMF model. The key advantage of this method is that the STNMF simultaneously compute sparse similarity matrix, clustering rows and columns of the input data matrix. Finally, our experiment results are presented.
展开▼
