English

Exact and Heuristic Algorithms for Semi-Nonnegative Matrix Factorization

Numerical Analysis 2015-10-28 v3 Machine Learning Numerical Analysis Optimization and Control Machine Learning

Abstract

Given a matrix MM (not necessarily nonnegative) and a factorization rank rr, semi-nonnegative matrix factorization (semi-NMF) looks for a matrix UU with rr columns and a nonnegative matrix VV with rr rows such that UVUV is the best possible approximation of MM according to some metric. In this paper, we study the properties of semi-NMF from which we develop exact and heuristic algorithms. Our contribution is threefold. First, we prove that the error of a semi-NMF of rank rr has to be smaller than the best unconstrained approximation of rank r1r-1. This leads us to a new initialization procedure based on the singular value decomposition (SVD) with a guarantee on the quality of the approximation. Second, we propose an exact algorithm (that is, an algorithm that finds an optimal solution), also based on the SVD, for a certain class of matrices (including nonnegative irreducible matrices) from which we derive an initialization for matrices not belonging to that class. Numerical experiments illustrate that this second approach performs extremely well, and allows us to compute optimal semi-NMF decompositions in many situations. Finally, we analyze the computational complexity of semi-NMF proving its NP-hardness, already in the rank-one case (that is, for r=1r = 1), and we show that semi-NMF is sometimes ill-posed (that is, an optimal solution does not exist).

Keywords

Cite

@article{arxiv.1410.7220,
  title  = {Exact and Heuristic Algorithms for Semi-Nonnegative Matrix Factorization},
  author = {Nicolas Gillis and Abhishek Kumar},
  journal= {arXiv preprint arXiv:1410.7220},
  year   = {2015}
}

Comments

22 pages, 6 figures. New: comparison with k-means initialization, numerical results on real data, ill-posedness of semi-NMF

R2 v1 2026-06-22T06:37:16.112Z