A Comparative Study of Iterative Solvers Exploiting Spectral Information for SPD Systems

@article{Giraud2006ACS,
  title={A Comparative Study of Iterative Solvers Exploiting Spectral Information for SPD Systems},
  author={Luc Giraud and Daniel Ruiz and Ahmed Touhami},
  journal={SIAM J. Sci. Comput.},
  year={2006},
  volume={27},
  pages={1760-1786}
}
When solving the symmetric positive definite (SPD) linear system ${\bf A} {\bf x}^\star = {\bf b}$ with the conjugate gradient method, the smallest eigenvalues in the matrix ${\bf A}$ often slow down the convergence. Consequently if the smallest eigenvalues in ${\bf A}$ could somehow be "removed," the convergence may be improved. This observation is of importance even when a preconditioner is used, and some extra techniques might be investigated to further improve the convergence rate of the… 
Krylov and Polynomial Iterative Solvers Combined with Partial Spectral Factorization for SPD Linear Systems
TLDR
This work compares the numerical efficiency, computational complexity, and sensitivity to the accuracy of the spectral information of the techniques presented in [1], [2] and [3].
Using partial spectral information for block diagonal preconditioning of saddle-point systems
Considering saddle-point systems of the Karush–Kuhn–Tucker (KKT) form, we propose approximations of the “ideal” block diagonal preconditioner based on the exact Schur complement proposed by Murphy et
Partial Spectral Information from Linear Systems to Speed-Up Numerical Simulations in Computational Fluid Dynamics
TLDR
This work proposes to perform one single computation of the eigenspace associated with the smallest eigenvalues, at the beginning of the iterative process, to improve the convergence of the Krylov method used in subsequent iterations of the fluid flow algorithm by means of this pre-computed partial spectral information.
Inexact Subspace Iteration to Accelerate the Solution of Linear Systems with Multiple Right-Hand Sides
TLDR
A two-phase acceleration technique for the solution of Symmetric and Positive Definite linear systems with multiple right-hand sides is proposed and validated as a good way to speed up the solutions of symmetric and positive definite linear systems.
A Hybrid Approach Combining Chebyshev Filter and Conjugate Gradient for Solving Linear Systems with Multiple Right-Hand Sides
TLDR
This work proposes an approach based on a combination of the conjugate gradient method with Chebyshev filtering polynomials, applied only to a part of the spectrum of the coefficient matrix, as preconditioners that target some specific convergence properties of the conjunction gradient method.
Inexact Subspace Iteration for the Consecutive Solution of Linear Systems with Changing Right-Hand Sides
TLDR
A two-phase acceleration technique for the solution of Symmetric and Positive Definite linear systems with multiple right-hand sides is proposed and it is shown that it is possible to determine when to stop the inner iteration in order to achieve the targeted invariance in the outer iteration.
Theoretical and numerical comparison of various projection methods derived from deflation, domain decomposition and multigrid methods
TLDR
A suggestion of the optimal second-level preconditionser, which is as stable as the abstract balancing preconditioner and as cheap and fast as the deflation preconditiouser is suggested.
Convergence analysis of the ChebFilterCG algorithm
TLDR
A convergence analysis of ChebFilterCG is presented, showing theoretically and numerically that the algorithm yields an approximation of the invariant subspace associated with the smallest eigenvalues that can be recycled for solving several linear systems with the same matrix and different right‐hand sides.
Fast Inexact Implicitly Restarted Arnoldi Method for Generalized Eigenvalue Problems with Spectral Transformation
  • Fei Xue, H. Elman
  • Mathematics, Computer Science
    SIAM J. Matrix Anal. Appl.
  • 2012
TLDR
A new tuning strategy constructed from vectors in both previous and current IRA cycles is studied, and it is shown how tuning is used in a new two-phase algorithm to greatly reduce inner iteration counts.
THEORETICAL AND NUMERICAL COMPARISON OF PROJECTED CG METHODS DERIVED FROM DEFLATION, DOMAIN DECOMPOSITION AND MULTIGRID METHODS
TLDR
A suggestion of a two-level preconditionser, that is as robust as the abstract balancing preconditioner and as cheap and fast as the deflation precONDitioner is suggested.
...
...

References

SHOWING 1-10 OF 45 REFERENCES
Krylov and Polynomial Iterative Solvers Combined with Partial Spectral Factorization for SPD Linear Systems
TLDR
This work compares the numerical efficiency, computational complexity, and sensitivity to the accuracy of the spectral information of the techniques presented in [1], [2] and [3].
A Class of Spectral Two-Level Preconditioners
TLDR
This paper proposes a class of preconditioners both for unsymmetric and for symmetric linear systems that can also be adapted for asymmetric positive definite problems and shows the advantages of the preconditionsers for solving dense linear systems arising in electromagnetism applications.
Analysis of Projection Methods for Solving Linear Systems with Multiple Right-Hand Sides
TLDR
A superconvergence behavior of the CG process of the seed system when compared with the usual CG process is observed and theoretical proof is given to justify these observations.
A Chebyshev-based two-stage iterative method as an alternative to the direct solution of linear systems
TLDR
Combining Chebyshev iterations with the block Lanczos algorithm, it is proposed a way to identify and extract precise information related to the ill-conditioned part of the given linear system, and to use it for the computation of extra solutionsrelated to the same matrix with changing right-hand sides.
Preconditioned conjugate gradients for solving singular systems
Adaptively Preconditioned GMRES Algorithms
TLDR
Two new methods are described for determining preconditioners from spectral information gathered by the Arnoldi process during iterations by the restarted GMRES algorithm to determine an invariant subspace of the matrix A associated with eigenvalues close to the origin and to move these eigen values so that a higher rate of convergence of the iterative methods is achieved.
Implicitly Restarted GMRES and Arnoldi Methods for Nonsymmetric Systems of Equations
  • R. Morgan
  • Computer Science
    SIAM J. Matrix Anal. Appl.
  • 2000
TLDR
This work presents algorithms that use implicit restarting in order to retain information that deflates the smallest eigenvalues and thus improves the convergence of the generalized minimum residual method.
An Augmented Conjugate Gradient Method for Solving Consecutive Symmetric Positive Definite Linear Systems
TLDR
New theoretical results for the conjugate gradient method are given and a new version is proposed that generates a Krylov subspace which can be efficiently recycled thanks to orthogonal projections in subsequent systems.
...
...