English

On Restricted Nonnegative Matrix Factorization

Formal Languages and Automata Theory 2016-05-24 v1 Computational Complexity Machine Learning

Abstract

Nonnegative matrix factorization (NMF) is the problem of decomposing a given nonnegative n×mn \times m matrix MM into a product of a nonnegative n×dn \times d matrix WW and a nonnegative d×md \times m matrix HH. Restricted NMF requires in addition that the column spaces of MM and WW coincide. Finding the minimal inner dimension dd is known to be NP-hard, both for NMF and restricted NMF. We show that restricted NMF is closely related to a question about the nature of minimal probabilistic automata, posed by Paz in his seminal 1971 textbook. We use this connection to answer Paz's question negatively, thus falsifying a positive answer claimed in 1974. Furthermore, we investigate whether a rational matrix MM always has a restricted NMF of minimal inner dimension whose factors WW and HH are also rational. We show that this holds for matrices MM of rank at most 33 and we exhibit a rank-44 matrix for which WW and HH require irrational entries.

Keywords

Cite

@article{arxiv.1605.07061,
  title  = {On Restricted Nonnegative Matrix Factorization},
  author = {Dmitry Chistikov and Stefan Kiefer and Ines Marušić and Mahsa Shirmohammadi and James Worrell},
  journal= {arXiv preprint arXiv:1605.07061},
  year   = {2016}
}

Comments

Full version of an ICALP'16 paper

R2 v1 2026-06-22T14:07:21.491Z