Effective and Robust Preconditioning of General SPD Matrices via Structured Incomplete Factorization

For general symmetric positive definite (SPD) matrices, we present a framework for designing effective and robust black-box preconditioners via structured incomplete factorization. In a scaling-and-compression strategy, off-diagonal blocks are first scaled on both sides (by the inverses of the factors of the corresponding diagonal blocks) and then compressed into low-rank approximations. ULV-type factorizations are then computed. A resulting prototype preconditioner is always positive definite. Generalizations to practical hierarchical multilevel preconditioners are given. Systematic analysis of the approximation error, robustness, and effectiveness is shown for both the prototype preconditioner and the multilevel generalization. In particular, we show how local scaling and compression control the approximation accuracy and robustness and how aggressive compression leads to efficient preconditioners that can significantly reduce the condition number and improve the eigenvalue clustering. A result is also given to show when the multilevel preconditioner preserves positive definiteness. The costs to apply the multilevel preconditioners are about O(N), where N is the matrix size. Numerical tests on several ill-conditioned problems show the effectiveness and robustness even if the compression uses very small numerical ranks. In addition, significant robustness and effectiveness benefits can be observed as compared with a standard rank structured preconditioner based on direct off-diagonal compression.

[1]  T. Manteuffel.  An incomplete factorization technique for positive definite linear systems , 1980 .

[2]  John K. Reid,et al.  The Multifrontal Solution of Indefinite Sparse Symmetric Linear , 1983, TOMS.

[3]  Michele Benzi,et al.  Robust Approximate Inverse Preconditioning for the Conjugate Gradient Method , 2000, SIAM J. Sci. Comput..

[4]  J. Meijerink,et al.  An iterative solution method for linear systems of which the coefficient matrix is a symmetric -matrix , 1977 .

[5]  Yousef Saad,et al.  An Algebraic Multilevel Preconditioner with Low-Rank Corrections for Sparse Symmetric Matrices , 2016, SIAM J. Matrix Anal. Appl..

[6]  M. A. Ajiz,et al.  A robust incomplete Choleski‐conjugate gradient algorithm , 1984 .

[7]  Shivkumar Chandrasekaran,et al.  A Fast ULV Decomposition Solver for Hierarchically Semiseparable Representations , 2006, SIAM J. Matrix Anal. Appl..

[8]  Yousef Saad,et al.  Low-Rank Correction Methods for Algebraic Domain Decomposition Preconditioners , 2015, SIAM J. Matrix Anal. Appl..

[9]  L. Grasedyck,et al.  Domain-decomposition Based ℌ-LU Preconditioners , 2007 .

[10]  Xiaoye S. Li,et al.  Direction-Preserving and Schur-Monotonic Semiseparable Approximations of Symmetric Positive Definite Matrices , 2009, SIAM J. Matrix Anal. Appl..

[11]  B. Engquist,et al.  Sweeping preconditioner for the Helmholtz equation: Hierarchical matrix representation , 2010, 1007.4290.

[12]  Igor E. Kaporin,et al.  High quality preconditioning of a general symmetric positive definite matrix based on its UTU + UTR + RTU-decomposition , 1998, Numer. Linear Algebra Appl..

[13]  Jianlin Xia,et al.  On the Complexity of Some Hierarchical Structured Matrix Algorithms , 2012, SIAM J. Matrix Anal. Appl..

[14]  Jianlin Xia,et al.  A Superfast Structured Solver for Toeplitz Linear Systems via Randomized Sampling , 2012, SIAM J. Matrix Anal. Appl..

[15]  Jianlin Xia,et al.  Robust Approximate Cholesky Factorization of Rank-Structured Symmetric Positive Definite Matrices , 2010, SIAM J. Matrix Anal. Appl..

[16]  Per-Gunnar Martinsson,et al.  A Fast Randomized Algorithm for Computing a Hierarchically Semiseparable Representation of a Matrix , 2011, SIAM J. Matrix Anal. Appl..

[17]  Lexing Ying,et al.  Fast construction of hierarchical matrix representation from matrix-vector multiplication , 2009, J. Comput. Phys..

[18]  John P. Boyd,et al.  Numerical experiments on the condition number of the interpolation matrices for radial basis functions , 2011 .

[19]  Yousef Saad,et al.  Divide and Conquer Low-Rank Preconditioners for Symmetric Matrices , 2013, SIAM J. Sci. Comput..

[20]  Z. Drmač,et al.  On the Perturbation of the Cholesky Factorization , 1994 .

[21]  Igor E. Kaporin,et al.  High quality preconditioning of a general symmetric positive definite matrix based on its U , 1998 .

[22]  Yousef Saad,et al.  Schur complement‐based domain decomposition preconditioners with low‐rank corrections , 2015, Numer. Linear Algebra Appl..

[23]  Eric Darve,et al.  An efficient preconditioner for the fast simulation of a 2D Stokes flow in porous media , 2016 .

[24]  Jianlin Xia,et al.  Fast algorithms for hierarchically semiseparable matrices , 2010, Numer. Linear Algebra Appl..

[25]  Cornelis V. M. van der Mee,et al.  A Method for Generating Infinite Positive Self-adjoint Test Matrices and Riesz Bases , 2005, SIAM J. Matrix Anal. Appl..

[26]  Artem Napov,et al.  Conditioning Analysis of Incomplete Cholesky Factorizations with Orthogonal Dropping , 2013, SIAM J. Matrix Anal. Appl..