On the stabilization of approximation procedures for a class of integral equations of the first kind with noisy right-hand sides (Q4339993)

Linear integral equations of first kind with inaccurate data given at a finite set of points are considered. Regularization is achieved by decomposing the problem into a well-posed and an ill-posed part. The well-posed part is handling between a space of (smooth) functions (in which the solution is supposed to be lying) and the range of a linear smoothing operator acting on it. This part is treated by a convergent discretization scheme. For the ill-posed part (which is between the range of the operator and the space of the data) an interpolation-projection method is proposed and the properties required for its working are discussed. It is shown how things can be made to fit well together so that there is optimal convergence of the approximate to the true solution for the data error tending to zero as the measurement grid and the discretization are refined. The author concludes his paper with some well chosen examples.