Journal article
Online Nonnegative Matrix Factorization With Outliers
IEEE transactions on signal processing, Vol.65(3), pp.555-570
02/01/2017
DOI: 10.1109/TSP.2016.2620967
Abstract
We propose a unified and systematic framework for performing online nonnegative matrix factorization in the presence of outliers. Our framework is particularly suited to large-scale data. We propose two solvers based on projected gradient descent and the alternating direction method of multipliers. We prove that the sequence of objective values converges almost surely by appealing to the quasi-martingale convergence theorem. We also show the sequence of learned dictionaries converges to the set of stationary points of the expected loss function almost surely. In addition, we extend our basic problem formulation to various settings with different constraints and regularizers. We also adapt the solvers and analyses to each setting. We perform extensive experiments on both synthetic and real datasets. These experiments demonstrate the computational efficiency and efficacy of our algorithms on tasks such as (parts-based) basis learning, image denoising, shadow removal, and foreground-background separation.
Details
- Title: Subtitle
- Online Nonnegative Matrix Factorization With Outliers
- Creators
- Renbo Zhao - National University of SingaporeVincent Y. F. Tan - National University of Singapore
- Resource Type
- Journal article
- Publication Details
- IEEE transactions on signal processing, Vol.65(3), pp.555-570
- DOI
- 10.1109/TSP.2016.2620967
- ISSN
- 1053-587X
- eISSN
- 1941-0476
- Publisher
- IEEE
- Number of pages
- 16
- Grant note
- R-263-000-C12-112 / Ministry of Education Academic Research Fund R-263-000-B37-133 / NUS Young Investigator Award
- Language
- English
- Date published
- 02/01/2017
- Academic Unit
- Business Analytics
- Record Identifier
- 9984446457602771
Metrics
3 Record Views