Abstract

As an important approach of clustering, Non-negative Matrix factorizations (NMF) algorithm develops quickly in recent years. We summarized them in this paper from two viewpoints. On the one hand, the analysis and induction of some representative NMF algorithms have been made from several aspects, such as the ideas of algorithm, key technology, advantage and disadvantage. On the other hand, the evolution and convergence properties of hybrid methods, based on both sparsity and smoothness constraints for the resulting incremental NMF (INMF), are discussed. As INMF outputs in specific contexts, the interpretability provided us the opportunities for the modification of INMF algorithms in the future work.

Paper Details