dqds with Aggressive Early Deflation

The dqds algorithm computes all the singular values of an $n \times n$ bidiagonal matrix to high relative accuracy in $O(n^2)$ cost. Its efficient implementation is now available as a LAPACK subroutine and is the preferred algorithm for this purpose. In this paper we incorporate into dqds a technique called aggressive early deflation, which has been applied successfully to the Hessenberg QR algorithm. Extensive numerical experiments show that aggressive early deflation often reduces the dqds runtime significantly. In addition, our theoretical analysis suggests that with aggressive early deflation, the performance of dqds is largely independent of the shift strategy. We confirm through experiments that the zero-shift version is often as fast as the shifted version. We give a detailed error analysis to prove that with our proposed deflation strategy, dqds computes all the singular values to high relative accuracy.

[1]  William B. Gragg,et al.  The QR algorithm for unitary Hessenberg matrices , 1986 .

[2]  Kazuo Murota,et al.  Superquadratic convergence of DLASQ for computing matrix singular values , 2010, J. Comput. Appl. Math..

[3]  Charles R. Johnson.  A Gersgorin-type lower bound for the smallest singular value , 1989 .

[4]  Suely Oliveira.  A new parallel chasing algorithm for transforming arrowhead matrices to tridiagonal form , 1998, Math. Comput..

[5]  Beresford N. Parlett,et al.  An implementation of the dqds algorithm (positive case) , 2000 .

[6]  G. W. Stewart,et al.  Matrix Algorithms: Volume 1, Basic Decompositions , 1998 .

[7]  Paul Willems,et al.  On MR3-type Algorithms for the Tridiagonal Symmetric Eigenproblem and the Bidiagonal SVD , 2018 .

[8]  B. Parlett,et al.  Accurate singular values and differential qd algorithms , 1994 .

[9]  James Demmel,et al.  Accurate Singular Values of Bidiagonal Matrices , 1990, SIAM J. Sci. Comput..

[10]  James Demmel,et al.  Algorithm 880: A testing infrastructure for symmetric tridiagonal eigensolvers , 2008, TOMS.

[11]  B. Parlett,et al.  Multiple representations to compute orthogonal eigenvectors of symmetric tridiagonal matrices , 2004 .

[12]  I. Dhillon.  Algorithm for the Symmetric Tridiagonal Eigenvalue/Eigenvector Problem , 1998 .

[13]  Roy Mathias,et al.  Quadratic Residual Bounds for the Hermitian Eigenvalue Problem , 1998 .

[14]  Yamamoto Yusaku,et al.  A Fully Pipelined Multishift QR Algorithm for Parallel Solution of Symmetric Tridiagonal Eigenproblems , 2008 .

[15]  Danny C. Sorensen,et al.  Deflation for Implicitly Restarted Arnoldi Methods , 1998 .

[16]  G. W. Stewart,et al.  Matrix algorithms , 1998 .

[17]  Ren-Cang Li,et al.  A note on eigenvalues of perturbed Hermitian matrices , 2005 .

[18]  B. Kågström,et al.  The Multishift QZ Algorithm with Aggressive Early Deflation ? , 2006 .

[19]  Beresford N. Parlett,et al.  Another orthogonal matrix , 2006 .

[20]  Kazuo Murota,et al.  On Convergence of the DQDS Algorithm for Singular Value Computation , 2008, SIAM J. Matrix Anal. Appl..

[21]  B. Parlett.  The Symmetric Eigenvalue Problem , 1981 .

[22]  Efstratios Gallopoulos,et al.  Computing smallest singular triplets with implicitly restarted Lanczos bidiagonalization , 2004, Applied Numerical Mathematics.

[23]  James Demmel,et al.  Applied Numerical Linear Algebra , 1997 .

[24]  Daniel Kressner,et al.  The Effect of Aggressive Early Deflation on the Convergence of the QR Algorithm , 2008, SIAM J. Matrix Anal. Appl..

[25]  James Demmel,et al.  On a Block Implementation of Hessenberg Multishift QR Iteration , 1989, Int. J. High Speed Comput..

[26]  Inderjit S. Dhillon,et al.  Orthogonal Eigenvectors and Relative Gaps , 2003, SIAM J. Matrix Anal. Appl..

[27]  Robert A. van de Geijn,et al.  Deferred Shifting Schemes for Parallel QR Methods , 1993, SIAM J. Matrix Anal. Appl..

[28]  Gene H. Golub,et al.  Calculation of Gauss quadrature rules , 1967, Milestones in Matrix Computation.

[29]  B. Parlett,et al.  Forward Instability of Tridiagonal QR , 1993, SIAM J. Matrix Anal. Appl..