The method of successive extrapolated iterated defect correction and its application to second kind Fredholm's integral equations

We describe an application of the principle of Iterated Defect Correction (IDeC) on the quadrature methods for the numerical solution of Fredholm's integral equations of the second kind. We also derive an asymptotic expansion for the global error in the solution produced by the IDeC method. Applying Richardson extrapolation repeatedly on the IDeC method, we present the technique of Successive Extrapolated Iterated Defect Correction (SEIDeC) and the resulting asymptotic expansion for the global error. Numerical tests confirm the asymptotic results and demonstrate the power of the IDeC method as well as the superiority of our method SEIDeC.     

Besov regularity for second order elliptic boundary value problems with variable coefficients

This paper is concerned with some theoretical foundations for adaptive numerical methods for elliptic boundary value problems. The approximation order that can be achieved by such an adaptive method is determined by certain Besov regularity of the weak solution. We study Besov regularity for second order elliptic problems in bounded domains in ℝ ^d . The investigations are based on intermediate Schauder estimates and on some potential theoretic framework. Moreover, we use characterizations of Besov spaces by wavelet expansions. This work has been supported by the Deutsche Forschungsgemeinschaft (Da 360/1-1)     

IDeC — Convergence independent of error asymptotics

The present paper is concerned with the method of Iterated Defect Correction (IDeC) for two-point boundary value problems. We investigate the contractive behaviour of the IDeC iteration in a completely discrete setting. Our results (which are a generalization of classical results based on asymptotic expansions of the discretization error) imply the stability of the collocation method which defines the fixed point of the IDeC iteration.     

New fifth‐order two‐derivative Runge‐Kutta methods with constant and frequency‐dependent coefficients

Two‐derivative Runge‐Kutta methods are Runge‐Kutta methods for problems of the form y′ = f(y) that include the second derivative y′′ = g(y) = f ′(y)f(y) and were developed in the work of Chan and Tsai. In this work, we consider explicit methods and construct a family of fifth‐order methods with three stages of the general case that use several evaluations of f and g per step. For problems with oscillatory solution and in the case that a good estimate of the dominant frequency is known, methods with frequency‐dependent coefficients are used; there are several procedures for constructing such methods. We give the general framework for the construction of methods with variable coefficients following the approach of Simos. We modify the above family to derive methods with frequency‐dependent coefficients following this approach as well as the approach given by Vanden Berghe. We provide numerical results to demonstrate the efficiency of the new methods using three test problems.     

Implications of the Choice of Quadrature Nodes for Picard Integral Deferred Corrections Methods for Ordinary Differential Equations

This paper concerns a class of deferred correction methods recently developed for initial value ordinary differential equations; such methods are based on a Picard integral form of the correction equation. These methods divide a given timestep [t_n,t_n+1] into substeps, and use function values computed at these substeps to approximate the Picard integral by means of a numerical quadrature. The main purpose of this paper is to present a detailed analysis of the implications of the location of quadrature nodes on the accuracy and stability of the overall method. Comparisons between Gauss-Legendre, Gauss-Lobatto, Gauss-Radau, and uniformly spaced points are presented. Also, for a given set of quadrature nodes, quadrature rules may be formulated that include or exclude function values computed at the left-hand endpoint t_n. Quadrature rules that do not depend on the left-hand endpoint (which are referred to as right-hand quadrature rules) are shown to lead to L(α)-stable implicit methods with α≈π/2. The semi-implicit analog of this property is also discussed. Numerical results suggest that the use of uniform quadrature nodes, as opposed to nodes based on Gaussian quadratures, does not significantly affect the stability or accuracy of these methods for orders less than ten. In contrast, a study of the reduction of order for stiff equations shows that when uniform quadrature nodes are used in conjunction with a right-hand quadrature rule, the form and extent of order-reduction changes considerably. Specifically, a reduction of order to is observed for uniform nodes as opposed to for non-uniform nodes, where Δt denotes the time step and ε a stiffness parameter such that ε→0 corresponds to the problem becoming increasingly stiff. AMS subject classification (2000) 65B05     

A Survey of Algebraic Multilevel Iteration (AMLI) Methods

For the solution by preconditioned conjugate gradient methods of symmetric positive definite equations as arising in boundary value problems we consider preconditioning methods of AMLI type. Particular attention is devoted to providing methods of optimal order of computational complexity which in addition promise to be robust, i.e. with a convergence rate which is bounded above independently of size of discretization parameter h, jumps in problem coefficients, and shape of finite elements or, equivalently, anisotropy of problem coefficients. In addition, the computational cost per iteration step must have optimal order.New results on upper bounds of one of the important parameters in the methods, the Cauchy—Bunyakowski—Schwarz constant are given and an algebraic method how to improve its value is presented.This revised version was published online in October 2005 with corrections to the Cover Date.     

Improving performance when solving high-order and mixed-order boundary value problems in ODEs

Solving high-order or mixed-order boundary value problems by general purpose software often requires the system to be first converted to a larger equivalent first-order system. The cost of solving such problems is generally O(m ³), where m is the dimension of the equivalent first-order system. In this paper, we show how to reduce this cost by exploiting the special structure the “equivalent” first-order system inherits from the original associated mixed-order system. This technique applies to a broad class of boundary value methods. We illustrate the potential benefits by considering in detail a general purpose Runge–Kutta method and a multiple shooting method. This revised version was published online in June 2006 with corrections to the Cover Date.     

A posteriori error estimates for mixed finite element approximations of elliptic problems

We derive residual based a posteriori error estimates of the flux in L ²-norm for a general class of mixed methods for elliptic problems. The estimate is applicable to standard mixed methods such as the Raviart–Thomas–Nedelec and Brezzi–Douglas–Marini elements, as well as stabilized methods such as the Galerkin-Least squares method. The element residual in the estimate employs an elementwise computable postprocessed approximation of the displacement which gives optimal order.     

Multilevel augmentation methods for differential equations

We develop multilevel augmentation methods for solving differential equations. We first establish a theoretical framework for convergence analysis of the boundary value problems of differential equations, and then construct multiscale orthonormal bases in H₀^m(0,1) spaces. Finally, the multilevel augmentation methods in conjunction with the multiscale orthonormal bases are applied to two-point boundary value problems of both second-order and fourth-order differential equations. Theoretical analysis and numerical tests show that these methods are computationally stable, efficient and accurate. Dedicated to Professor Charles A. Micchelli on the occasion of his 60th birthday with friendship and esteem. Mathematics subject classifications (2000) 65J15, 65R20. Zhongying Chen: Supported in part by the Natural Science Foundation of China under grants 10371137 and 10201034, the Foundation of Doctoral Program of National Higher Education of China under grant 20030558008, Guangdong Provincial Natural Science Foundation of China under grant 1011170 and the Foundation of Zhongshan University Advanced Research Center. Yuesheng Xu: Corresponding author. Supported in part by the US National Science Foundation under grants 9973427 and 0312113, by NASA under grant NCC5-399, by the Natural Science Foundation of China under grant 10371122 and by the Chinese Academy of Sciences under the program of “One Hundred Distinguished Young Scientists”.     

Gradient methods for multiple state optimal design problems

We optimise a distribution of two isotropic materials α I and β I (α < β) occupying the given body in R ^d . The optimality is described by an integral functional (cost) depending on temperatures u ₁, . . . , u _m of the body obtained for different source terms f ₁, . . . ,f _m with homogeneous Dirichlet boundary conditions. The relaxation of this optimal design problem with multiple state equations is needed, introducing the notion of composite materials as fine mixtures of different phases, mathematically described by the homogenisation theory. The necessary conditions of optimality are derived via the Gateaux derivative of the cost functional. Unfortunately, there could exist points in which necessary conditions of optimality do not give any information on the optimal design. In the case m < d we show that there exists an optimal design which is a rank-m sequential laminate with matrix material α I almost everywhere on Ω. Contrary to the optimality criteria method, which is commonly used for the numerical solution of optimal design problems (although it does not rely on a firm theory of convergence), this result enables us to effectively use classical gradient methods for minimising the cost functional.      

Algebraically stable general linear methods and the <Emphasis Type="Italic">G</Emphasis>-matrix

The standard algebraic stability condition for general linear methods (GLMs) is considered in a modified form, and connected to a branch of Control Theory concerned with the discrete algebraic Riccati equation (DARE). The DARE theory shows that, for an algebraically stable method, there is a minimal G-matrix, G ^*, satisfying an equation, rather than an inequality. This result, and another alternative reformulation of algebraic stability, are applied to construct new GLMs with 2 steps and 2 stages, one of which has order p=4 and stage order q=3. The construction process is simplified by method-equivalence, and Butcher’s simplified order conditions for the case p≤q+1.      

Entropic proximal decomposition methods for convex programs and variational inequalities

We consider convex optimization and variational inequality problems with a given separable structure. We propose a new decomposition method for these problems which combines the recent logarithmic-quadratic proximal theory introduced by the authors with a decomposition method given by Chen-Teboulle for convex problems with particular structure. The resulting method allows to produce for the first time provably convergent decomposition schemes based on C ^∞ Lagrangians for solving convex structured problems. Under the only assumption that the primal-dual problems have nonempty solution sets, global convergence of the primal-dual sequences produced by the algorithm is established. Received: October 6, 1999 / Accepted: February 2001?Published online September 17, 2001     

Detecting “dense” columns in interior point methods for linear programs

During the iterations of interior point methods symmetric indefinite systems are decomposed by LD̂L ^T factorization. This step can be performed in a special way where the symmetric indefinite system is transformed to a positive definite one, called the normal equations system. This approach proved to be efficient in most of the cases and numerically reliable, due to the positive definite property. It has been recognized, however, that in case the linear program contains “dense” columns, this approach results in an undesirable fill–in during the computations and the direct factorization of the symmetric indefinite system is more advantageous. The paper describes a new approach to detect cases where the system of normal equations is not preferable for interior point methods and presents a new algorithm for detecting the set of columns which is responsible for the excessive fill–in in the matrix AA ^T . By solving large–scale linear programming problems we demonstrate that our heuristic is reliable in practice. This work was supported in part by the Hungarian Scientific Research Fund OTKA K60480.     

One-leg variable-coefficient formulas for ordinary differential equations and local–global step size control

In this paper we discuss a class of numerical algorithms termed one-leg methods. This concept was introduced by Dahlquist in 1975 with the purpose of studying nonlinear stability properties of multistep methods for ordinary differential equations. Later, it was found out that these methods are themselves suitable for numerical integration because of good stability. Here, we investigate one-leg formulas on nonuniform grids. We prove that there exist zero-stable one-leg variable-coefficient methods at least up to order 11 and give examples of two-step methods of orders 2 and 3. In this paper we also develop local and global error estimation techniques for one-leg methods and implement them with the local–global step size selection suggested by Kulikov and Shindin in 1999. The goal of this error control is to obtain automatically numerical solutions for any reasonable accuracy set by the user. We show that the error control is more complicated in one-leg methods, especially when applied to stiff problems. Thus, we adapt our local–global step size selection strategy to one-leg methods.     

Shohat-Favard and Chebyshev’s methods in d-orthogonality

This paper is concerned with the Shohat-Favard, Chebyshev and Modified Chebyshev methods for d-orthogonal polynomial sequences d∈ℕ. Shohat-Favard’s method is presented from the concept of dual sequence of a sequence of polynomials. We deduce the recurrence relations for the Chebyshev and the Modified Chebyshev methods for every d∈ℕ. The three methods are implemented in the Mathematica programming language. We show several formal and numerical tests realized with the software developed. This revised version was published online in June 2006 with corrections to the Cover Date.     

On the local convergence of adjoint Broyden methods

The numerical solution of nonlinear equation systems is often achieved by so-called quasi-Newton methods. They preserve the rapid local convergence of Newton’s method at a significantly reduced cost per step by successively approximating the system Jacobian though low-rank updates. We analyze two variants of the recently proposed adjoint Broyden update, which for the first time combines the classical least change property with heredity on affine systems. However, the new update does require, the evaluation of so-called adjoint vectors, namely products of the transposed Jacobian with certain dual direction vectors. The resulting quasi-Newton method is linear contravariant in the sense of Deuflhard (Newton methods for nonlinear equations. Springer, Heidelberg, 2006) and it is shown here to be locally and q-superlinearly convergent. Our numerical results on a range of test problems demonstrate that the new method usually outperforms Newton’s and Broyden’s method in terms of runtime and iterations count, respectively. Partially supported by the DFG Research Center Matheon “Mathematics for Key Technologies”, Berlin and the DFG grant WA 1607/2-1.     

On a Family of Gradient-Type Projection Methods for Nonlinear Ill-Posed Problems

Abstract

We propose and analyze a family of successive projection methods whose step direction is the same as the Landweber method for solving nonlinear ill-posed problems that satisfy the Tangential Cone Condition (TCC). This family encompasses the Landweber method, the minimal error method, and the steepest descent method; thus, providing an unified framework for the analysis of these methods. Moreover, we define new methods in this family, which are convergent for the constant of the TCC in a range twice as large as the one required for the Landweber and other gradient type methods. The TCC is widely used in the analysis of iterative methods for solving nonlinear ill-posed problems. The key idea in this work is to use the TCC in order to construct special convex sets possessing a separation property, and to successively project onto these sets. Numerical experiments are presented for a nonlinear two-dimensional elliptic parameter identification problem, validating the efficiency of our method.     

On the worst case complexity of potential reduction algorithms for linear programming

There are several classes of interior point algorithms that solve linear programming problems in O(√nL) iterations. Among them, several potential reduction algorithms combine both theoretical (O(√nL) iterations) and practical efficiency as they allow the flexibility of line searches in the potential function, and thus can lead to practical implementations. It is a significant open question whether interior point algorithms can lead to better complexity bounds. In the present paper we give some negative answers to this question for the class of potential reduction algorithms. We show that, even if we allow line searches in the potential function, and even for problems that have network structure, the bound O(√nL) is tight for several potential reduction algorithms, i.e., there is a class of examples with network structure, in which the algorithms need at least Θ(√nL) iterations to find an optimal solution. The research of the author was partially supported by a Presidential Young Investigator Award DDM-9158118 with matching funds from Draper Laboratory.     

A fixed-center spherical separation algorithm with kernel transformations for classification problems

We consider a special case of the optimal separation, via a sphere, of two discrete point sets in a finite dimensional Euclidean space. In fact we assume that the center of the sphere is fixed. In this case the problem reduces to the minimization of a convex and nonsmooth function of just one variable, which can be solved by means of an “ad hoc” method in O(p log p) time, where p is the dataset size. The approach is suitable for use in connection with kernel transformations of the type adopted in the support vector machine (SVM) approach. Despite of its simplicity the method has provided interesting results on several standard test problems drawn from the binary classification literature. This research has been partially supported by the Italian “Ministero dell’Istruzione, dell’Università e della Ricerca Scientifica”, under PRIN project Numerical Methods for Global Optimization and for some classes of Nonsmooth Optimization Problems (2005017083.002).