A Non-negative Matrix Factorization approach to state estimation in stochastic systems

In this paper a novel non-negative matrix factorization (NMF) based state estimation approach is applied to a stochastic system. Recognizing that uniqueness of solutions is a challenge in NMF in general, we analyze in the paper under what conditions NMF has a unique solution in the stochastic system state estimation context. Moreover, we explore the system attributes corresponding to those conditions. Examples are presented to illustrate the analysis and to manifest the effectiveness of the proposed algorithm.