Algorithms for nonnegative matrix and tensor factorizations: a unified view based on block coordinate descent framework

We review algorithms developed for nonnegativematrix factorization (NMF) and 1 nonnegative tensor factorization (NTF) from a unified view based on the block coordinate 2 descent (BCD) framework. NMF and NTF are low-rank approximation methods for matri3 ces and tensors in which the low-rank factors are constrained to have only nonnegative 4 elements. The nonnegativity constraints have been shown to enable natural interpretations 5 and allow better solutions in numerous applications including text analysis, computer vision, 6 and bioinformatics. However, the computation of NMF and NTF remains challenging and 7 expensive due the constraints. Numerous algorithmic approaches have been proposed to effi8 ciently compute NMF and NTF. The BCD framework in constrained non-linear optimization 9 readily explains the theoretical convergence properties of several efficient NMF and NTF 10 algorithms, which are consistent with experimental observations reported in literature. In 11 addition, we discuss algorithms that do not fit in the BCD framework contrasting them from 12 those based on the BCD framework. With insights acquired from the unified perspective, 13 we also propose efficient algorithms for updating NMF when there is a small change in the 14 reduced dimension or in the data. The effectiveness of the proposed updating algorithms are 15 validated experimentally with synthetic and real-world data sets. 16