NMF (Non-Negative Matrix Factorization)
Technique in multivariate analysis that factors high-dimensional vectors into a lower-dimensional representation, while preserving the non-negative elements in the data sets.
Non-Negative Matrix Factorization (NMF) is a group of algorithms in multivariate analysis where a matrix V is factorized into two matrices W and H, with the property that all three matrices have no negative elements. This non-negativity contributes to the interpretability of the parts-based representations. NMF finds applications in fields such as computer vision, text mining, pattern recognition, and bioinformatics. It is particularly useful when dealing with data where all attributes are of the same kind, and where zero is a meaningful number, representing the absence of quantity.
The concept of Non-Negative Matrix Factorization was introduced in 1999 by researchers Daniel D. Lee and H. Sebastian Seung, who illustrated the usefulness of NMF in the context of data compression and face recognition.
The primary contributors to the development of NMF are Daniel D. Lee and H. Sebastian Seung, who introduced the technique. They continue to push its limits, inventing newer, more efficient algorithms and extending its range of applications.