[Next]:  Integral equation methods for problems of  
 [Up]:  Project descriptions  
 [Previous]:  Optimization problems with random and nonsmooth  
 [Contents]   [Index] 


Regularization of severely ill-posed problems and applications

Collaborator: G. Bruckner

Cooperation with: J. Cheng (Fudan University, Shanghai, China), S.V. Pereverzev (National Academy of Sciences of Ukraine, Kiev)

Description: In the mathematical treatment of inverse problems ranging from tomography over non-destructive testing to satellite geodetic exploration, operator equations of the form

Ax = y      (1)
arise, where A is a linear operator, and the right-hand side y is given only approximately.

The problems (1) are ill-posed, in most cases even severely ill-posed, and have to be regularized for solution. Since the problems are usually given in infinite-dimensional spaces they must be additionally discretized. Therefore the well-known method of regularization by discretization seems to be natural and is frequently used. Moreover, because of the lack of information about the singular values of A and the smoothness of the solution x, for constructing regularization procedures, the a posteriori parameter choice by discrepancy principles has been proposed. Recently, T. Hohage [3] considered discrepancy principles with respect to Tikhonov's regularization, while the investigations of B. Kaltenbacher [4] referred to the moderately ill-posed case.

In the present project, the idea of regularization by discretization is continued with a new strategy of a posteriori parameter choice that assumes only estimates for the singular values and the smoothness of the solution (cf. [2]). The algorithm is not based on a discrepancy principle and selects a relevant discretized solution from a number of discretized solutions for subsequent discretization levels. The selected approximate solution has an optimal order of accuracy. Moreover, the algorithm is applied to a problem with a logarithmic convolution-type operator (cf. [1]), for which the assumptions can be verified. Further work is needed to apply this method to nondestructive testing, quality control of grating devices or other real-world problems.

References:

  1. G. BRUCKNER, J. CHENG, Tikhonov regularization for an integral equation of the first kind with logarithmic kernel, J. Inverse and Ill-posed Problems, 8 (2000), pp. 665-675.
  2. G. BRUCKNER, S.V. PEREVERZEV, Self-regularization of projection methods with a posteriori discretization level choice for severely ill-posed problems, to appear in: Inverse Problems.
  3. T. HOHAGE, Regularization of exponentially ill-posed problems, Numer. Funct. Anal. Optim., 21 (2000), pp. 439-464.
  4. B. KALTENBACHER, Regularization by projection with a posteriori discretization level choice for linear and nonlinear ill-posed problems, Inverse Problems, 16 (2000), pp. 1523-1539.



 [Next]:  Integral equation methods for problems of  
 [Up]:  Project descriptions  
 [Previous]:  Optimization problems with random and nonsmooth  
 [Contents]   [Index] 

LaTeX typesetting by I. Bremer
5/16/2003