Sergei V. Pereverzev

On the Adaptive Selection of the Parameter in Regularization of Ill-Posed Problems

We study the possibility of using the structure of the regularization error for a posteriori choice of the regularization parameter. As a result, a rather general form of a selection criterion is proposed, and its relation to the heuristical quasi-optimality principle of Tikhonov and Glasko [Z. Vychisl. Mat. Mat. Fiz., 4 (1964), pp. 564-571] and to an adaptation scheme proposed in a statistical context by Lepskii [Theory Probab. Appl., 36 (1990), pp. 454-466] is discussed. The advantages of the proposed criterion are illustrated by using such examples as self-regularization of the trapezoidal rule for noisy Abel-type integral equations, Lavrentiev regularization for nonlinear ill-posed problems, and an inverse problem of the two-dimensional profile reconstruction.

Regularization of some linear ill-posed problems with discretized random noisy data

For linear statistical ill-posed problems in Hilbert spaces we introduce an adaptive procedure to recover the unknown so- lution from indirect discrete and noisy data.. This procedure is shown to be order optimal for a large class of problems. Smooth- ness of the solution is measured in terms of general source condi- tions. The concept of operator monotone functions turns out to be an important tool for the analysis.

Regularization in Hilbert scales under general smoothing conditions

For solving linear ill-posed problems regularization methods are required when the available data include some noise. In the present paper regularized approximations are obtained by a general regularization scheme in Hilbert scales which include well-known regularization methods such as the method of Tikhonov regularization and its higher-order forms, spectral methods, asymptotical regularization and iterative regularization methods. For both the cases of high- and low-order regularization, we study a priori and a posteriori rules for choosing the regularization parameter and provide order optimal error bounds that characterize the accuracy of the regularized approximations. These error bounds have been obtained under general smoothing conditions. The results extend earlier results and cover the case of finitely and infinitely smoothing operators. The theory is illustrated by a special ill-posed deconvolution problem arising in geoscience.

Discretization strategy for linear ill-posed problems in variable Hilbert scales

The authors study the regularization of projection methods for solving linear ill-posed problems with compact and injective linear operators in Hilbert spaces. The smoothness of the unknown solution is given in terms of general source conditions, such that the framework of variable Hilbert scales is suitable. The structure of the error is analysed in terms of the noise level, the regularization parameter and as a function of other parameters, driving the discretization. As a result, a strategy is proposed which automatically adapts to the unknown source condition, uniformly for certain classes, and provides the optimal order of accuracy.

Optimal Discretization of Inverse Problems in Hilbert Scales. Regularization and Self-Regularization of Projection Methods

We study the efficiency of the approximate solution of ill-posed problems, based on discretized noisy observations, which we assume to be given beforehand. A basic purpose of the paper is the consideration of stochastic noise, but deterministic noise is also briefly discussed. We restrict ourselves to problems which can be formulated in Hilbert scales. Within this framework we shall quantify the degree of ill-posedness, provide general conditions on projection schemes to achieve the best possible order of accuracy. We pay particular attention on the problem of self-regularization vs. Tikhonov regularization. Moreover, we study the information complexity. Asymptotically, any method which achieves the best possible order of accuracy must use at least such amount of noisy observations.

A Carleman estimate and the balancing principle in the quasi-reversibility method for solving the Cauchy problem for the Laplace equation

The quasi-reversibility method of solving the Cauchy problem for the Laplace equation in a bounded domain ? is considered. With the help of the Carleman estimation technique improved error and stability bounds in a subdomain ?? ? ? are obtained. This paves the way for the use of the balancing principle for an a posteriori choice of the regularization parameter ? in the quasi-reversibility method. As an adaptive regularization parameter choice strategy, the balancing principle does not require a priori knowledge of either the solution smoothness or a constant K appearing in the stability bound estimation. Nevertheless, this principle allows an a posteriori parameter choice that up to a controllable constant achieves the best accuracy guaranteed by the Carleman estimate.

Multi-parameter regularization and its numerical realization

In this paper we propose and analyse a choice of parameters in the multi-parameter regularization of Tikhonov type. A modified discrepancy principle is presented within the multi-parameter regularization framework. An order optimal error bound is obtained under the standard smoothness assumptions. We also propose a numerical realization of the multi-parameter discrepancy principle based on the model function approximation. Numerical experiments on a series of test problems support theoretical results. Finally we show how the proposed approach can be successfully implemented in Laplacian Regularized Least Squares for learning from labeled and unlabeled examples.

Numerical differentiation from a viewpoint of regularization theory

In this paper, we discuss the classical ill-posed problem of numerical differentiation, assuming that the smoothness of the function to be differentiated is unknown. Using recent results on adaptive regularization of general ill-posed problems, we propose new rules for the choice of the stepsize in the finite-difference methods, and for the regularization parameter choice in numerical differentiation regularized by the iterated Tikhonov method. These methods are shown to be effective for the differentiation of noisy functions, and the order-optimal convergence results for them are proved.

Information Complexity of Multivariate Fredholm Integral Equations in Sobolev Classes

In this paper, the complexity of full solution of Fredholm integral equations of the second kind with data from the Sobolev classWr2is studied. The exact order of information complexity is derived. The lower bound is proved using a Gelfand number technique. The upper bound is shown by providing a concrete algorithm of optimal order, based on a specific hyperbolic cross approximation of the kernel function. Numerical experiments are included, comparing the optimal algorithm with the standard Galerkin method.

Regularization without preliminary knowledge of smoothness and error behaviour

The mathematical formulation of many physical problems results in the task of inverting a compact operator. The only known sensible solution technique is regularization which poses a severe problem in itself. Classically one dealt with deterministic noise models and required both the knowledge of smoothness of the solution function and the overall error behavior. We will show that we can guarantee an asymptotically optimal regularization for a physically motivated noise model under no assumptions for the smoothness and rather weak assumptions on the noise behavior which can mostly obtained out of two input data sets. An application to the determination of the gravitational field out of satellite data will be shown.