Abstract:Linear systems in applications are typically well-posed, and yet the coefficient matrices may be nearly singular in that the condition number $\kappa(\boldsymbol{A})$ may be close to $1/\varepsilon_{w}$, where $\varepsilon_{w}$ denotes the unit roundoff of the working precision. It is well known that iterative refinement (IR) can make the forward error independent of $\kappa(\boldsymbol{A})$ if $\kappa(\boldsymbol{A})$ is sufficiently smaller than $1/\varepsilon_{w}$ and the residual is computed in higher precision. We propose a new iterative method, called Forward-and-Backward Stabilized Minimal Residual or FBSMR, by conceptually hybridizing right-preconditioned GMRES (RP-GMRES) with quasi-minimization. We develop FBSMR based on a new theoretical framework of essential-forward-and-backward stability (EFBS), which extends the backward error analysis to consider the intrinsic condition number of a well-posed problem. We stabilize the forward and backward errors in RP-GMRES to achieve EFBS by evaluating a small portion of the algorithm in higher precision while evaluating the preconditioner in lower precision. FBSMR can achieve optimal accuracy in terms of both forward and backward errors for well-posed problems with unpolluted matrices, independently of $\kappa(\boldsymbol{A})$. With low-precision preconditioning, FBSMR can reduce the computational, memory, and energy requirements over direct methods with or without IR. FBSMR can also leverage parallelization-friendly classical Gram-Schmidt in Arnoldi iterations without compromising EFBS. We demonstrate the effectiveness of FBSMR using both random and realistic linear systems.

Restarted Weighted Full Orthogonalization Method for Shifted Linear Systems

Restarted weighted full orthogonalization method for shifted linear systems

Restarted Hessenberg Method for Solving Shifted Nonsymmetric Linear Systems

Restarted Hessenberg Method for Shifted Nonsymmetric Linear Systems with Applications to Fractional Differential Equations

Global GMRES with Deflated Restarting for Families of Shifted Linear Systems

On restarted and deflated block FOM and GMRES methods for sequences of shifted linear systems

Solving Shifted Linear Systems with Restarted GMRES Augmented with Error Approximations.

A Parallel Reduced-Space Interior Point Method With Orthogonal Collocation for First-Swing Stability Constrained Emergency Control

Near-optimal convergence of the full orthogonalization method

A flexible and adaptive Simpler GMRES with deflated restarting for shifted linear systems

Structure Preserving Balanced Proper Orthogonal Decomposition for Second-Order Form Systems Via Shifted Legendre Polynomials

Restarted Global FOM and GMRES Algorithms for the Stein-like Matrix Equationx+m(x)=c

Optimal Solutions of Well-Posed Linear Systems via Low-Precision Right-Preconditioned GMRES with Forward and Backward Stabilization

Orthogonal Weight Normalization: Solution to Optimization over Multiple Dependent Stiefel Manifolds in Deep Neural Networks

On IOM(q): the Incomplete Orthogonalization Method for Large Unsymmetric Linear Systems

A Refined Second-order Arnoldi (RSOAR) Method for the Quadratic Eigenvalue Problem and Implicit Restarting

A Block GMRES Method with Deflated Restarting for Solving Linear Systems with Multiple Shifts and Multiple Right-Hand Sides

A NEW ALGORITHM FOR THE COMPUTATION OF LOW FREQUENCY ELECTRO-MECHANICAL OSCILLATION MODES OF LARGE POWER SYSTEMS

A Simpler GMRES and Its Adaptive Variant for Shifted Linear Systems.

Global FOM and GMRES algorithms for a class of complex matrix equations.

On IGMRES: an Incomplete Generalized Minimal Residual Method for Large Unsymmetric Linear Systems