BLENDENPIK: SUPERCHARGING LAPACK'S LEAST-SQUARES SOLVER

被引:138
作者
Avron, Haim [1 ]
Maymounkov, Petar [2 ]
Toledo, Sivan [1 ]
机构
[1] Tel Aviv Univ, Raymond & Beverly Sackler Fac Exact Sci, Blavatnik Sch Comp Sci, IL-69978 Tel Aviv, Israel
[2] MIT, Comp Sci & Artificial Intelligence Lab, Cambridge, MA 02139 USA
基金
以色列科学基金会;
关键词
dense linear least squares; randomized numerical linear algebra; randomized pre-conditioners; RANDOMIZED ALGORITHM;
D O I
10.1137/090767911
中图分类号
O29 [应用数学];
学科分类号
070104 ;
摘要
Several innovative random-sampling and random-mixing techniques for solving problems in linear algebra have been proposed in the last decade, but they have not yet made a significant impact on numerical linear algebra. We show that by using a high-quality implementation of one of these techniques, we obtain a solver that performs extremely well in the traditional yardsticks of numerical linear algebra: it is significantly faster than high-performance implementations of existing state-of-the-art algorithms, and it is numerically backward stable. More specifically, we describe a least-squares solver for dense highly overdetermined systems that achieves residuals similar to those of direct QR factorization-based solvers (lapack), outperforms lapack by large factors, and scales significantly better than any QR-based solver.
引用
收藏
页码:1217 / 1236
页数:20
相关论文
共 26 条
[1]  
Ailon N, 2008, PROCEEDINGS OF THE NINETEENTH ANNUAL ACM-SIAM SYMPOSIUM ON DISCRETE ALGORITHMS, P1
[2]  
Ailon Nir, 2006, P 38 ANN ACM S THEOR, P557, DOI [DOI 10.1145/1132516.1132597, 10.1145/1132516.1132597]
[3]  
[Anonymous], ABS07101435 CORR
[4]  
Avron H., 2010, THESIS TEL AVIV U IS
[5]   USING PERTURBED QR FACTORIZATIONS TO SOLVE LINEAR LEAST-SQUARES PROBLEMS [J].
Avron, Haim ;
Ng, Esmond ;
Toledo, Sivan .
SIAM JOURNAL ON MATRIX ANALYSIS AND APPLICATIONS, 2009, 31 (02) :674-693
[6]   Random projections for the nonnegative least-squares problem [J].
Boutsidis, Christos ;
Drineas, Petros .
LINEAR ALGEBRA AND ITS APPLICATIONS, 2009, 431 (5-7) :760-771
[7]  
CANDES EJ, 2008, ABS08054471 CORR
[8]   STOPPING CRITERIA FOR THE ITERATIVE SOLUTION OF LINEAR LEAST SQUARES PROBLEMS [J].
Chang, X. -W. ;
Paige, C. C. ;
Titley-Peloquin, D. .
SIAM JOURNAL ON MATRIX ANALYSIS AND APPLICATIONS, 2009, 31 (02) :831-852
[9]   SAMPLING ALGORITHMS AND CORESETS FOR lp REGRESSION [J].
Dasgupta, Anirban ;
Drineas, Petros ;
Harb, Boulos ;
Kumar, Ravi ;
Mahoney, Michael W. .
SIAM JOURNAL ON COMPUTING, 2009, 38 (05) :2060-2078
[10]   Fast linear algebra is stable [J].
Demmel, James ;
Dumitriu, Ioana ;
Holtz, Olga .
NUMERISCHE MATHEMATIK, 2007, 108 (01) :59-91