Two algorithms for orthogonal nonnegative matrix factorization with application to clustering精品.pdfVIP

Neurocomputing 141 (2014) 15–25 Contents lists available at ScienceDirect Neurocomputing journal homepage: /locate/neucom Two algorithms for orthogonal nonnegative matrix factorization with application to clustering $ Filippo Pompili a, Nicolas Gillis b,n, P.-A. Absil c, François Glineur c,d a Department of Electronic and Information Engineering, University of Perugia, Italy b Department of Mathematics and Operational Research, Université de Mons, Rue de Houdain 9, 7000 Mons, Belgium c Université catholique de Louvain, ICTEAM Institute, B-1348 Louvain-la-Neuve, Belgium d Université catholique de Louvain, CORE, Voie du Roman Pays 34, B-1348 Louvain-la-Neuve, Belgium a r t i c l e i n f o a b s t r a c t Article history: Approximate matrix factorization techniques with both nonnegativity and orthogonality constraints, Received 21 June 2013 referred to as orthogonal nonnegative matrix factorization (ONMF), have been recently introduced and Received in revised form shown to work remarkably well for clustering tasks such as document classification. In this paper, we 24 January 2014 introduce two new methods to solve ONMF. First, we show mathematical equivalence between ONMF Accepted 3 February 2014 and a weighted variant of spherical k-means, from which we derive our first method, a simple EM-like Available online 8 April 2014 algorithm. This also allows us to determine when ONMF should