Randomized Projection for Rank-Revealing Matrix Factorizations and Low-Rank Approximations
Abstract
Rank-revealing matrix decompositions provide an essential tool in spectral analysis of matrices, including the Singular Value Decomposition (SVD) and related low-rank approximation techniques. QR with Column Pivoting (QRCP) is usually suitable for these purposes, but it can be much slower than the unpivoted QR algorithm. For large matrices, the difference in performance is due to increased communication between the processor and slow memory, which QRCP needs in order to choose pivots during decomposition. Our main algorithm, Randomized QR with Column Pivoting (RQRCP), uses randomized projection to make pivot decisions from a much smaller sample matrix, which we can construct to reside in a faster level of memory than the original matrix. This technique may be understood as trading vastly reduced communication for a controlled increase in uncertainty during the decision process. Furthermore, for rank-revealing purposes, the selection mechanism in RQRCP produces results that are the same quality as the standard algorithm, but with performance near that of unpivoted QR (often an order of magnitude faster for large matrices). Additionally, we also propose two formulas that facilitate further performance improvements. The first efficiently updates sample matrices to avoid computing new randomized projections. The second avoids large trailing updates during the decompositionmore »
- Authors:
-
- Sandia National Lab. (SNL-CA), Livermore, CA (United States)
- Univ. of California, Berkeley, CA (United States)
- Publication Date:
- Research Org.:
- Sandia National Lab. (SNL-CA), Livermore, CA (United States)
- Sponsoring Org.:
- USDOE National Nuclear Security Administration (NNSA); National Science Foundation (NSF)
- OSTI Identifier:
- 1650159
- Report Number(s):
- SAND-2020-8281J
Journal ID: ISSN 0036-1445; 689858
- Grant/Contract Number:
- AC04-94AL85000; NA0003525; 1319312; 1760316
- Resource Type:
- Accepted Manuscript
- Journal Name:
- SIAM Review
- Additional Journal Information:
- Journal Volume: 62; Journal Issue: 3; Journal ID: ISSN 0036-1445
- Publisher:
- Society for Industrial and Applied Mathematics
- Country of Publication:
- United States
- Language:
- English
- Subject:
- 97 MATHEMATICS AND COMPUTING; QR factorization; column pivoting; rank-revealing; random sampling; sample update; blocked algorithm; low-rank approximation; truncated SVD
Citation Formats
Duersch, Jed A., and Gu, Ming. Randomized Projection for Rank-Revealing Matrix Factorizations and Low-Rank Approximations. United States: N. p., 2020.
Web. doi:10.1137/20m1335571.
Duersch, Jed A., & Gu, Ming. Randomized Projection for Rank-Revealing Matrix Factorizations and Low-Rank Approximations. United States. https://doi.org/10.1137/20m1335571
Duersch, Jed A., and Gu, Ming. Thu .
"Randomized Projection for Rank-Revealing Matrix Factorizations and Low-Rank Approximations". United States. https://doi.org/10.1137/20m1335571. https://www.osti.gov/servlets/purl/1650159.
@article{osti_1650159,
title = {Randomized Projection for Rank-Revealing Matrix Factorizations and Low-Rank Approximations},
author = {Duersch, Jed A. and Gu, Ming},
abstractNote = {Rank-revealing matrix decompositions provide an essential tool in spectral analysis of matrices, including the Singular Value Decomposition (SVD) and related low-rank approximation techniques. QR with Column Pivoting (QRCP) is usually suitable for these purposes, but it can be much slower than the unpivoted QR algorithm. For large matrices, the difference in performance is due to increased communication between the processor and slow memory, which QRCP needs in order to choose pivots during decomposition. Our main algorithm, Randomized QR with Column Pivoting (RQRCP), uses randomized projection to make pivot decisions from a much smaller sample matrix, which we can construct to reside in a faster level of memory than the original matrix. This technique may be understood as trading vastly reduced communication for a controlled increase in uncertainty during the decision process. Furthermore, for rank-revealing purposes, the selection mechanism in RQRCP produces results that are the same quality as the standard algorithm, but with performance near that of unpivoted QR (often an order of magnitude faster for large matrices). Additionally, we also propose two formulas that facilitate further performance improvements. The first efficiently updates sample matrices to avoid computing new randomized projections. The second avoids large trailing updates during the decomposition in truncated low-rank approximations. Our truncated version of RQRCP also provides a key initial step in our truncated SVD approximation, TUXV. These advances open up a new performance domain for large matrix factorizations that will support efficient problem-solving techniques for challenging applications in science, engineering, and data analysis.},
doi = {10.1137/20m1335571},
journal = {SIAM Review},
number = 3,
volume = 62,
place = {United States},
year = {Thu Aug 06 00:00:00 EDT 2020},
month = {Thu Aug 06 00:00:00 EDT 2020}
}
Works referenced in this record:
A Practical Randomized CP Tensor Decomposition
journal, January 2018
- Battaglino, Casey; Ballard, Grey; Kolda, Tamara G.
- SIAM Journal on Matrix Analysis and Applications, Vol. 39, Issue 2
The WY Representation for Products of Householder Matrices
journal, January 1987
- Bischof, Christian; Van Loan, Charles
- SIAM Journal on Scientific and Statistical Computing, Vol. 8, Issue 1
A Parallel QR Factorization Algorithm with Controlled Local Pivoting
journal, January 1991
- Bischof, Christian H.
- SIAM Journal on Scientific and Statistical Computing, Vol. 12, Issue 1
Structure-Preserving and Rank-Revealing QR-Factorizations
journal, November 1991
- Bischof, Christian H.; Hansen, Per Christian
- SIAM Journal on Scientific and Statistical Computing, Vol. 12, Issue 6
Rank revealing QR factorizations
journal, April 1987
- Chan, Tony F.
- Linear Algebra and its Applications, Vol. 88-89
Some Applications of the Rank Revealing QR Factorization
journal, May 1992
- Chan, Tony F.; Hansen, Per Christian
- SIAM Journal on Scientific and Statistical Computing, Vol. 13, Issue 3
Communication-optimal Parallel and Sequential QR and LU Factorizations
journal, January 2012
- Demmel, James; Grigori, Laura; Hoemmen, Mark
- SIAM Journal on Scientific Computing, Vol. 34, Issue 1
Communication Avoiding Rank Revealing QR Factorization with Column Pivoting
journal, January 2015
- Demmel, James W.; Grigori, Laura; Gu, Ming
- SIAM Journal on Matrix Analysis and Applications, Vol. 36, Issue 1
Randomized QR with Column Pivoting
journal, January 2017
- Duersch, Jed A.; Gu, Ming
- SIAM Journal on Scientific Computing, Vol. 39, Issue 4
Randomized Matrix Decompositions Using R
journal, January 2019
- Erichson, N. Benjamin; Voronin, Sergey; Brunton, Steven L.
- Journal of Statistical Software, Vol. 89, Issue 11
Efficient Algorithms for Computing a Strong Rank-Revealing QR Factorization
journal, July 1996
- Gu, Ming; Eisenstat, Stanley C.
- SIAM Journal on Scientific Computing, Vol. 17, Issue 4
Finding Structure with Randomness: Probabilistic Algorithms for Constructing Approximate Matrix Decompositions
journal, January 2011
- Halko, N.; Martinsson, P. G.; Tropp, J. A.
- SIAM Review, Vol. 53, Issue 2
Basis selection in LOBPCG
journal, October 2006
- Hetmaniuk, U.; Lehoucq, R.
- Journal of Computational Physics, Vol. 218, Issue 1
QR factorization with complete pivoting and accurate computation of the SVD
journal, April 2000
- Higham, Nicholas J.
- Linear Algebra and its Applications, Vol. 309, Issue 1-3
Generalized Canonical Polyadic Tensor Decomposition
journal, January 2020
- Hong, David; Kolda, Tamara G.; Duersch, Jed A.
- SIAM Review, Vol. 62, Issue 1
The Selection of Prior Distributions by Formal Rules
journal, September 1996
- Kass, Robert E.; Wasserman, Larry
- Journal of the American Statistical Association, Vol. 91, Issue 435
Randomized algorithms for the low-rank approximation of matrices
journal, December 2007
- Liberty, E.; Woolfe, F.; Martinsson, P. -G.
- Proceedings of the National Academy of Sciences, Vol. 104, Issue 51
randUTV: A Blocked Randomized Algorithm for Computing a Rank-Revealing UTV Factorization
journal, March 2019
- Martinsson, P. G.; Quintana-Ortí, G.; Heavner, N.
- ACM Transactions on Mathematical Software, Vol. 45, Issue 1
Householder QR Factorization With Randomization for Column Pivoting (HQRRP)
journal, January 2017
- Martinsson, Per-Gunnar; Quintana OrtÍ, Gregorio; Heavner, Nathan
- SIAM Journal on Scientific Computing, Vol. 39, Issue 2
A randomized algorithm for the decomposition of matrices
journal, January 2011
- Martinsson, Per-Gunnar; Rokhlin, Vladimir; Tygert, Mark
- Applied and Computational Harmonic Analysis, Vol. 30, Issue 1
A Randomized Blocked Algorithm for Efficiently Computing Rank-revealing Factorizations of Matrices
journal, January 2016
- Martinsson, Per-Gunnar; Voronin, Sergey
- SIAM Journal on Scientific Computing, Vol. 38, Issue 5
Modification of the Householder Method Based on the Compact WY Representation
journal, May 1992
- Puglisi, Chiara
- SIAM Journal on Scientific and Statistical Computing, Vol. 13, Issue 3
A BLAS-3 Version of the QR Factorization with Column Pivoting
journal, September 1998
- Quintana-Ortí, Gregorio; Sun, Xiaobai; Bischof, Christian H.
- SIAM Journal on Scientific Computing, Vol. 19, Issue 5
A Randomized Algorithm for Principal Component Analysis
journal, January 2010
- Rokhlin, Vladimir; Szlam, Arthur; Tygert, Mark
- SIAM Journal on Matrix Analysis and Applications, Vol. 31, Issue 3
A Storage-Efficient $WY$ Representation for Products of Householder Transformations
journal, January 1989
- Schreiber, Robert; Van Loan, Charles
- SIAM Journal on Scientific and Statistical Computing, Vol. 10, Issue 1
A Block Orthogonalization Procedure with Constant Synchronization Requirements
journal, January 2002
- Stathopoulos, Andreas; Wu, Kesheng
- SIAM Journal on Scientific Computing, Vol. 23, Issue 6
The QLP Approximation to the Singular Value Decomposition
journal, January 1999
- Stewart, G. W.
- SIAM Journal on Scientific Computing, Vol. 20, Issue 4
A fast randomized algorithm for the approximation of matrices
journal, November 2008
- Woolfe, Franco; Liberty, Edo; Rokhlin, Vladimir
- Applied and Computational Harmonic Analysis, Vol. 25, Issue 3
A randomized algorithm for the decomposition of matrices
journal, January 2011
- Martinsson, Per-Gunnar; Rokhlin, Vladimir; Tygert, Mark
- Applied and Computational Harmonic Analysis, Vol. 30, Issue 1
Basis selection in LOBPCG
journal, October 2006
- Hetmaniuk, U.; Lehoucq, R.
- Journal of Computational Physics, Vol. 218, Issue 1
Randomized algorithms for the low-rank approximation of matrices
journal, December 2007
- Liberty, E.; Woolfe, F.; Martinsson, P. -G.
- Proceedings of the National Academy of Sciences, Vol. 104, Issue 51
Extensions of Lipschitz mappings into a Hilbert space
book, January 1984
- Johnson, William B.; Lindenstrauss, Joram
- Conference on Modern Analysis and Probability
Modification of the Householder Method Based on the Compact WY Representation
journal, May 1992
- Puglisi, Chiara
- SIAM Journal on Scientific and Statistical Computing, Vol. 13, Issue 3
Efficient Algorithms for Computing a Strong Rank-Revealing QR Factorization
journal, July 1996
- Gu, Ming; Eisenstat, Stanley C.
- SIAM Journal on Scientific Computing, Vol. 17, Issue 4
Finding structure with randomness: Probabilistic algorithms for constructing approximate matrix decompositions
text, January 2009
- Halko, Nathan; Martinsson, Per-Gunnar; Tropp, Joel A.
- arXiv
randUTV: A blocked randomized algorithm for computing a rank-revealing UTV factorization
preprint, January 2017
- Martinsson, Per-Gunnar; Quintana-Orti, Gregorio; Heavner, Nathan
- arXiv
Randomized Numerical Linear Algebra: Foundations & Algorithms
preprint, January 2020
- Martinsson, Per-Gunnar; Tropp, Joel
- arXiv