Weighted FOM and GMRES for solving nonsymmetric linear systems

This paper presents two new methods called WFOM and WGMRES, which are variants of FOM and GMRES, for solving large and sparse nonsymmetric linear systems. To accelerate the convergence, these new methods use a different inner product instead of the Euclidean one. Furthermore, at each restart, a different inner product is chosen. The weighted Arnoldi process is introduced for implementing these methods. After describing the weighted methods, we give the relations that link them to FOM and GMRES. Experimental results are presented to show the good performances of the new methods compared to FOM(m) and GMRES(m). This revised version was published online in August 2006 with corrections to the Cover Date.     

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

Numerical Algorithms - The problem of shifted linear systems is an important and challenging issue in a number of research applications. Krylov subspace methods are effective techniques for...     

Iterative versus direct parallel substructuring methods in semiconductor device modelling

The numerical simulation of semiconductor devices is extremely demanding in term of computational time because it involves complex embedded numerical schemes. At the kernel of these schemes is the solution of very ill‐conditioned large linear systems. In this paper, we present the various ingredients of some hybrid iterative schemes that play a central role in the robustness of these solvers when they are embedded in other numerical procedures. On a set of two‐dimensional unstructured mixed finite element problems representative of semiconductor simulation, we perform a fair and detailed comparison between parallel iterative and direct linear solution techniques. We show that iterative solvers can be robust enough to solve the very challenging linear systems that arise in those simulations. Copyright © 2004 John Wiley & Sons, Ltd.     

BiCR variants of the hybrid BiCG methods for solving linear systems with nonsymmetric matrices

We propose Bi-Conjugate Residual (BiCR) variants of the hybrid Bi-Conjugate Gradient (BiCG) methods (referred to as the hybrid BiCR variants) for solving linear systems with nonsymmetric coefficient matrices. The recurrence formulas used to update an approximation and a residual vector are the same as those used in the corresponding hybrid BiCG method, but the recurrence coefficients are different; they are determined so as to compute the coefficients of the residual polynomial of BiCR. From our experience it appears that the hybrid BiCR variants often converge faster than their BiCG counterpart. Numerical experiments show that our proposed hybrid BiCR variants are more effective and less affected by rounding errors. The factor in the loss of convergence speed is analyzed to clarify the difference of the convergence between our proposed hybrid BiCR variants and the hybrid BiCG methods.     

Iterative methods for solving linear equations

This paper presents some of the original versions of the conjugate-gradient method for solving a system of linear equations of the formAx=k.This paper originally appeared as NAML Report No. 52-9, 1951. Its preparation was supported in part by the Office of Naval Research.     

Iterative methods for convection dominated flow

Some iterative methods are considered for the numerical solution of convection diffusion problems. The first class of iterative methods is Chebyshev accelerated iterations. The issues of parameter selection and convergence rates are considered. Secondly, we consider convection—diffusion type iterations where the iterations are of Peaceman-Rachford type. Here, a conjecture is given concerning a related problem in functional analysis. Finally, we consider flow-directed iterative schemes. We describe some schemes of this class for an upwind difference method, and also for a nonlinear hyperbolic equation. We emphasize work that remains to be done on these methods.     

Iterative methods for overflow queuing models II

Summary Preconditioned conjugate gradient methods are employed to find the steady-state probability distribution of Markovian queuing networks that have overflow capacity. Different singular preconditioners that can be handled by separation of variables are discussed. The resulting preconditioned systems are nonsingular. Numerical results show that the number of iterations required for convergence grows very slowly with the queue sizes.This research was supported in part by the National Science Foundation grant DCR-8405506 and DCR-8602563     

Iterative parameter identification methods for nonlinear functions

This paper considers identification problems of nonlinear functions fitting or nonlinear systems modelling. A gradient based iterative algorithm and a Newton iterative algorithm are presented to determine the parameters of a nonlinear system by using the negative gradient search method and Newton method. Furthermore, two model transformation based iterative methods are proposed in order to enhance computational efficiencies. By means of the model transformation, a simpler nonlinear model is achieved to simplify the computation. Finally, the proposed approaches are analyzed using a numerical example.     

Iterative methods for overflow queueing models I

Summary Markovian queueing networks having overflow capacity are discussed. The Kolmogorov balance equations result in a linear homogeneous system, where the right null-vector is the steady-state probability distribution for the network. Preconditioned conjugate gradient methods are employed to find the null-vector. The preconditioner is a singular matrix which can be handled by separation of variables. The resulting preconditioned system is nonsingular. Numerical results show that the number of iterations required for convergence is roughly constant independent of the queue sizes. Analytic results are given to explain this fast convergence.     

Iterative methods for variational and complementarity problems

In this paper, we study both the local and global convergence of various iterative methods for solving the variational inequality and the nonlinear complementarity problems. Included among such methods are the Newton and several successive overrelaxation algorithms. For the most part, the study is concerned with the family of linear approximation methods. These are iterative methods in which a sequence of vectors is generated by solving certain linearized subproblems. Convergence to a solution of the given variational or complementarity problem is established by using three different yet related approaches. The paper also studies a special class of variational inequality problems arising from such applications as computing traffic and economic spatial equilibria. Finally, several convergence results are obtained for some nonlinear approximation methods.This research was based on work supported by the National Science Foundation under grant ECS-7926320.     

Iterative methods for synthesis of optical elements

Translated from Matematicheskie Modeli i Optimizatsiya Vychislitel'nykh Algoritmov, pp. 144–150, 1993.     

Iterative methods for solving general quasi-variational inequalities

In this paper, we introduce a new class of variational inequalities, which is called the general quasi-variational inequality. We establish the equivalence among the general quasi variational inequality and implicit fixed point problems and the Wiener–Hopf equations. We use this equivalent formulation to discuss the existence of a solution of the general quasi-variational inequality. This equivalent formulation is used to suggest and analyze some iterative algorithms for solving the general quasi-variational inequality. We also discuss the convergence analysis of these iterative methods. Several special cases are also discussed.     

Iterative methods for solving fredholm integral equations

A Gauss-Seidel type of iterative method is described for solving the non-linear Fredholm integral equation. The analysis shows that this method may be expected to converge faster than the standard iterative method.     

Iterative splitting methods for solving time-dependent problems

There is increasing motivation for solving time-dependent differential equations with iterative splitting schemes. While Magnus expansion has been intensively studied and widely applied for solving explicitly time-dependent problems, the combination with iterative splitting schemes can open up new areas. The main problems with the Magnus expansion are the exponential character and the difficulty of deriving practical higher order algorithms. An alternative method is based on iterative splitting methods that take into account a temporally inhomogeneous equation. In this work, we show that the ideas derived from the iterative splitting methods can be used to solve time-dependent problems. Examples are discussed.     

Iterative methods for stabilized discrete convection-diffusion problems

In this paper we study the computational cost of solving theconvection-diffusion equation using various discretization strategiesand iteration solution algorithms. The choice of discretizationinfluences the properties of the discrete solution and alsothe choice of solution algorithm. The discretizations consideredhere are stabilized low-order finite element schemes using streamlinediffusion, crosswind diffusion and shock-capturing. The latter,shock-capturing discretizations lead to nonlinear algebraicsystems and require nonlinear algorithms. We compare variouspreconditioned Krylov subspace methods including Newton-Krylovmethods for nonlinear problems, as well as several preconditionersbased on relaxation and incomplete factorization. We find thatalthough enhanced stabilization based on shock-capturing requiresfewer degrees of freedom than linear stabilizations to achievecomparable accuracy, the nonlinear algebraic systems are morecostly to solve than those derived from a judicious combinationof streamline diffusion and crosswind diffusion. Solution algorithmsbased on GMRES with incomplete block-matrix factorization preconditioningare robust and efficient.     

Preconditioned GMRES methods with incomplete Givens orthogonalization method for large sparse least-squares problems

We propose to precondition the GMRES method by using the incomplete Givens orthogonalization (IGO) method for the solution of large sparse linear least-squares problems. Theoretical analysis shows that the preconditioner satisfies the sufficient condition that can guarantee that the preconditioned GMRES method will never break down and always give the least-squares solution of the original problem. Numerical experiments further confirm that the new preconditioner is efficient. We also find that the IGO preconditioned BA-GMRES method is superior to the corresponding CGLS method for ill-conditioned and singular least-squares problems.     

Iterative selection methods for common fixed point problems

Many problems encountered in applied mathematics can be recast as the problem of selecting a particular common fixed point of a countable family of quasi-nonexpansive operators in a Hilbert space. We propose two iterative methods to solve such problems. Our convergence analysis is shown to cover a variety of existing results in this area. Applications to solving monotone inclusion and equilibrium problems are considered.     

Iterative methods for solving proximal split minimization problems

In this paper, we propose two iterative algorithms for finding the minimum-norm solution of a split minimization problem. We prove strong convergence of the sequences generated by the proposed algorithms. The iterative schemes are proposed in such a way that the selection of the step-sizes does not need any prior information about the operator norm. We further give some examples to numerically verify the efficiency and implementation of our new methods and compare the two algorithms presented. Our results act as supplements to several recent important results in this area.     

Iterative methods for a class of convolution equations

We suggest a method for constructing iterative parallel algorithms for integral convolution equations.     

Iterative methods for ill-posed problems and semiconvergent sequences

Iterative schemes, such as LSQR and RRGMRES, are among the most efficient methods for the solution of large-scale ill-posed problems. The iterates generated by these methods form semiconvergent sequences. A meaningful approximation of the desired solution of an ill-posed problem often can be obtained by choosing a suitable member of this sequence. However, it is not always a simple matter to decide which member to choose. Semiconvergent sequences also arise when approximating integrals by asymptotic expansions, and considerable experience and analysis of how to choose a suitable member of a semiconvergent sequence in this context are available. The present note explores how the guidelines developed within the context of asymptotic expansions can be applied to iterative methods for ill-posed problems.