Solution of finite systems of equations by interval iteration

In actual practice, iteration methods applied to the solution of finite systems of equations yield inconclusive results as to the existence or nonexistence of solutions and the accuracy of any approximate solutions obtained. On the other hand, construction of interval extensions of ordinary iteration operators permits one to carry out interval iteration computationally, with results which can give rigorous guarantees of existence or nonexistence of solutions, and error bounds for approximate solutions. Examples are given of the solution of a nonlinear system of equations and the calculation of eigenvalues and eigenvectors of a matrix by interval iteration. Several ways to obtain lower and upper bounds for eigenvalues are given.Sponsored by the United States Army under Contract No. DAAG29-80-C-0041.     

基于向量值有理插值的最优预测算法研究及陶瓷需求量预测的应用

通过对高维数据整体表达式建模预测方法和分区间等预测算法的缺陷分析,提出基于向量值有理插值的最优预测算法,通过有理向量插值函数和各分量的误差限得到向量之间的相似性,克服了其它很多算法利用向量的整体表达式方法而产生预测的偏差;另外,通过向量的误差限与训练样本所得向量值有理插值函数及迭代仿真方法来确定预测样本向量所对应的最优预测值.通过实例,算法所得预测值的精度比其他算法更高,并且分析了误差限和迭代步长对算法性能的影响.     

Interval methods for fixed-point problems

Interval analysis is applied to the fixed-point problem x=?(x) for continuous ?:S→S, where the space S is constructed from Cartesian products of the set R of real numbers, with componentwise definitions of arithmetic operations, ordering, and the product topology. With the aid of an interval inclusion φ:IS → IS in the interval space IS corresponding to S, interval iteration is used to establish the existence or nonexistence of a fixed point x^? of ? in the initial interval X₀. Each step of the interval iteration provides lower and upper bounds for fixed points of ? in the initial interval, from which approximate values and guaranteed error bounds can be obtained directly. In addition to interval iteration, operator equation and dissection methods are considered briefly.

The theory of interval iteration applies directly when only finite subsets of S, IS are used, so this method is adaptable immediately to actual computation. A numerical example is given of the use of interval iteration for the computational solution of a nonlinear integral equation of radiative transfer. It is shown that numerical results with acceptable, guaranteed accuracy can be obtained with a modest amount of computation for an extended range of the parameter involved.     

Error bounds for stochastic shortest path problems

For stochastic shortest path problems, error bounds for value iteration due to Bertsekas elegantly generalize the classic MacQueen–Porteus error bounds for discounted infinite-horizon Markov decision problems, but incur prohibitive computational overhead. We derive bounds on these error bounds that can be computed with little or no overhead, making them useful in practice—especially so, since easily-computed error bounds have not previously been available for this class of problems.     

Computable bounds for eigenvalues and eigenfunctions of elliptic differential operators

Summary We are concerned with bounds for the error between given approximations and the exact eigenvalues and eigenfunctions of self-adjoint operators in Hilbert spaces. The case is included where the approximations of the eigenfunctions don't belong to the domain of definition of the operator. For the eigenvalue problem with symmetric elliptic differential operators these bounds cover the case where the trial functions don't satisfy the boundary conditions of the problem. The error bounds suggest a certain defectminization method for solving the eigenvalue problems. The method is applied to the membrane problem.     

On the convergence of operator splitting for the Rosenau–Burgers equation

We present convergence analysis of operator splitting methods applied to the nonlinear Rosenau–Burgers equation. The equation is first splitted into an unbounded linear part and a bounded nonlinear part and then operator splitting methods of Lie‐Trotter and Strang type are applied to the equation. The local error bounds are obtained by using an approach based on the differential theory of operators in Banach space and error terms of one and two‐dimensional numerical quadratures via Lie commutator bounds. The global error estimates are obtained via a Lady Windermere's fan argument. Lastly, a numerical example is studied to confirm the expected convergence order.     

Error bounds in the approximation of eigenvalues of differential and integral operators

Various methods of approximating the eigenvalues and invariant subspaces of nonself-adjoint differential and integral operators are unified in a general theory. Error bounds are given, from which most of the error bounds in the literature can be derived. Computable error bounds are given for simple eigenvalues, and trace formulae are used to improve the accuracy of the computed eigenvalues.     

A note on application of integral operator in learning theory

By the aid of the properties of the square root of positive operators we refine the consistency analysis of regularized least square regression in a reproducing kernel Hilbert space. Sharper error bounds and faster learning rates are obtained when the sampling sequence satisfies a strongly mixing condition.     

Numerical methods for the linear Boltzmann transport equation in slab geometry

An iterative method for computing numerical solutions of a finite-difference system corresponding to the linear Boltzmann equation in slab geometry is presented. This iterative scheme gives a straightforward marching process starting from the given boundary and initial conditions. It is shown that with a suitable initial iteration the sequence of iterations converges monotonically to a unique solution of the finite-difference system. This monotone convergence leads to improved upper and lower bounds of the solution in each iteration, and to the well-posedness of the discrete system in the sense of Hadamard. It also leads to the convergence of the discrete system to the continuous system as the mesh size of the space–velocity–time variables approaches to zero. Under a mild restriction on the time-increment the discrete system is numerically stable, independent of the mesh-size of the space and velocity. An error estimate for the computed solution due to simultaneous initial and iteration error is obtained. Also given are some numerical results for the time-dependent and the steady-state solutions.     

Newton's method for nonlinear ordinary and partial differential equations

A unified approach is presented for proving the local, uniform and quadratic convergence of the approximate solutions and a-posteriori error bounds obtained by Newton's method for systems of nonlinear ordinary or partial differential equations satisfying an inverse-positive property. An important step is to show that, at each iteration, the linearized problem is inverse-positive. Many classes of problems are shown to satisfy this property. The convergence proofs depend crucially on an error bound derived previously by Rosen and the author for quasilinear elliptic, parabolic and hyperbolic problems.     

具一致广义Lipschitz连续算子的带误差的多步迭代间的收敛等价性

研究了一致光滑Banach空间中具一致广义Lipschitz连续的逐次渐近Φ-强伪压缩型算子的具误差的修正Mann迭代和具误差的修正多步Noor迭代间的收敛等价性问题,所得结果是对2007年Zhenyu Huang在一致光滑Banach空间中所建立的逼近具有有界值域的逐次Φ-强伪压缩算子的不动点具误差的修正Mann迭代和具误差的修正Ishikawa迭代两者的收敛是等价的这一结论更本质的和更一般的推广,所用的方法不全同于ZhenyuHuang所使用的方法,因此,从更一般的意义上肯定地回答了Rhoades和Soltuz于2003年所提出的猜想.     

Convergence order estimates of meshless collocation methods using radial basis functions

We study meshless collocation methods using radial basis functions to approximate regular solutions of systems of equations with linear differential or integral operators. Our method can be interpreted as one of the emerging meshless methods, cf. T. Belytschko et al. (1996). Its range of application is not confined to elliptic problems. However, the application to the boundary value problem for an elliptic operator, connected with an integral equation, is given as an example. Although the method has been used for special cases for about ten years, cf. E.J. Kansa (1990), there are no error bounds known. We put the main emphasis on detailed proofs of such error bounds, following the general outline described in C. Franke and R. Schaback (preprint). This revised version was published online in June 2006 with corrections to the Cover Date.     

Completely continuous and related multilinear operators

Completely continuous multilinear operators are defined and their properties investigated. This class of operators is shown to form a closed multi-ideal. Unlike the linear case, compact multilinear operators need not be completely continuous. The completely continuous maps are shown to be the closure of a subspace of the finite rank operators. Hilbert-Schmidt operators are also considered. An application to finding error bounds for solutions of multipower equations is presented.     

On the discrete spectrum of non-selfadjoint operators

We prove quantitative bounds on the eigenvalues of non-selfadjoint unbounded operators obtained from selfadjoint operators by a perturbation that is relatively-Schatten. These bounds are applied to obtain new results on the distribution of eigenvalues of Schrödinger operators with complex potentials.     

k-次增生与k-次散逸算子方程带误差的迭代序列收敛率的估计

在任意实Banach空间中,研究了Lipschitz的k-次增生算子方程x+Tx=f和k-次散逸算子方程x-λTx=f的解的带误差的收敛性与稳定性问题,并给出了收敛率的估计式,从而在很大程度上统一和发展了有关文献中的相应结果.     

On Newton’s method for solving equations containing Fréchet-differentiable operators of order at least two

We provide sufficient conditions for the semilocal convergence of Newton’s method to a locally unique solution of a nonlinear operator equation containing operators that are Fréchet-differentiable of order at least two, in a Banach space setting. Numerical examples are also provided to show that our results apply to solve nonlinear equations in cases earlier ones cannot [J.M. Gutiérrez, A new semilocal convergence theorem for Newton’s method, J. Comput. Appl. Math. 79(1997) 131-145; Z. Huang, A note of Kantorovich theorem for Newton iteration, J. Comput. Appl. Math. 47 (1993) 211-217; F.A. Potra, Sharp error bounds for a class of Newton-like methods, Libertas Mathematica 5 (1985) 71-84].     

A family of univariate rational Bernstein operators

We define and study a new family of univariate rational Bernstein operators. They are positive operators exact on linear polynomials. Moreover, like classical polynomial Bernstein operators, they enjoy the traditional shape preserving properties and they are total variation diminishing. Finally, for a specific class of denominators, some convergence results are proved, in particular a Voronovskaja theorem, and some error bounds are given.     

Efficient algorithms for robust and stable principal component pursuit problems

The problem of recovering a low-rank matrix from a set of observations corrupted with gross sparse error is known as the robust principal component analysis (RPCA) and has many applications in computer vision, image processing and web data ranking. It has been shown that under certain conditions, the solution to the NP-hard RPCA problem can be obtained by solving a convex optimization problem, namely the robust principal component pursuit (RPCP). Moreover, if the observed data matrix has also been corrupted by a dense noise matrix in addition to gross sparse error, then the stable principal component pursuit (SPCP) problem is solved to recover the low-rank matrix. In this paper, we develop efficient algorithms with provable iteration complexity bounds for solving RPCP and SPCP. Numerical results on problems with millions of variables and constraints such as foreground extraction from surveillance video, shadow and specularity removal from face images and video denoising from heavily corrupted data show that our algorithms are competitive to current state-of-the-art solvers for RPCP and SPCP in terms of accuracy and speed.     

B-矩阵线性互补问题的误差界估计

利用严格对角占优M-矩阵的逆矩阵的无穷大范数的范围,给出了B-矩阵线性互补问题误差界新的估计式.相应数值算例表明了结果的有效性.     

A fixed point technique to refine a simple approximate eigenvalue and a corresponding eigenvector

Let T be a bounded operator on a Banach space X. Let λ₀ be a nonzero simple eigenvalue of a ‘nearby’ operator T₀ and let ?₀ be a corresponding eigenvector. Several modified versions of a fixed point scheme are given for iteratively refining the initial approximations λ₀ and ?₀ of an eigenvalue λ of T and a corresponding eigenvector ? Convergence of these schemes is proved by considering error bounds for the iterates. These bounds hold if a compact operator T is approximated in the norm or in a Collectively compact manner by a sequence (T₀) of bounded operators, and λ₀ and ?₀ are eigenelements of T_n0 for a fixed _n0 of ‘moderate’ size. Numerical examples are no included to illustrate the performation of various iteration schemes.