Sequential randomized matrix factorization

Author

Bopardikar, Shaunak D. ; Nair, S.S. ; Rai, Rajesh

Author_Institution

United Technol. Res. Center, Inc., Berkeley, CA, USA

fYear

2013

fDate

17-19 June 2013

Firstpage

2705

Lastpage

2710

Abstract

Matrix factorization techniques play a key role in numerous data analysis and scientific computing tasks. This work extends recent research which uses randomized algorithms for performing matrix factorization.We develop and analyze a randomized method for incremental matrix factorization under certain uniform low-rankedness assumption. The main objective of this paper is to make the case for sequential randomized algorithms for approximate matrix factorizations of very large scale matrices. Numerical results and rigorous analysis of error bounds within a formal framework for the proposed sequential randomized algorithms are also outlined.

Keywords

data analysis; learning (artificial intelligence); matrix decomposition; randomised algorithms; data analysis; dimension reduction; scientific computing tasks; sequential randomized algorithms; sequential randomized matrix factorization technique; uniform low-rankedness assumption; very large scale matrices; Algorithm design and analysis; Approximation algorithms; Approximation methods; Computational complexity; Matrix decomposition; Sparse matrices; Vectors; Low-rank Decomposition; Matrix Factorization; Randomization;

fLanguage

English

Publisher

ieee

Conference_Titel

American Control Conference (ACC), 2013

Conference_Location

Washington, DC

ISSN

0743-1619

Print_ISBN

978-1-4799-0177-7

Type

conf

DOI

10.1109/ACC.2013.6580243

Filename

6580243