Potluru, Vamsi K. and Le Roux, Jonathan and Calhoun, Vince D. and Plis, Sergey M. and Pearlmutter, Barak A. and Hayes, Thoms P. (2013) Block Coordinate Descent for Sparse NMF. In: ICLR 2013 (First International Conference on Learning Representations), 2-4 May 2013, Scottsdale Arizona, USA.

Preview

Download (390kB) | Preview

Abstract

Nonnegative matrix factorization (NMF) has become a ubiquitous tool for data analysis. An important variant is the sparse NMF problem which arises when we explicitly require the learnt features to be sparse. A natural measure of sparsity is the La norm, however its optimization is NP-hard. Mixed norms, such as Ll/L2 measure, have been shown to model sparsity robustly, based on intuitive attributes that such measures need to satisfy. This is in contrast to computationally cheaper alternatives such as the plain L1 norm. However, present algorithms designed for optimizing the mixed norm L1 /L2 are slow and other formulations for sparse NMF have been proposed such as those based on L1 and La norms. Our proposed algorithm allows us to solve the mixed norm sparsity constraints while not sacrificing computation time. We present experimental evidence on real-world datasets that shows our new algorithm performs an order of magnitude faster compared to the current state-of-the-art solvers optimizing the mixed norm and is suitable for large-scale datasets.

Item Type:	Conference or Workshop Item (Paper)
Keywords:	Block Coordinate Descent; Sparse NMF; Nonnegative matrix factorization;
Academic Unit:	Faculty of Science and Engineering > Computer Science
Item ID:	6556
Depositing User:	Barak Pearlmutter
Date Deposited:	10 Nov 2015 16:41
Refereed:	Yes
URI:
Use Licence:	This item is available under a Creative Commons Attribution Non Commercial Share Alike Licence (CC BY-NC-SA). Details of this licence are available here

Repository Staff Only(login required)

Item control page

Downloads

Origin of downloads