Iterative regularization techniques for ill-posed problems (Q1921704)

First this work presents a survey of recent results on iterative regularization techniques (i.e. strongly convergent and perturbation stable) for solving unstable problems. Afterwards several new results concerning the study of monotone processes for ill-posed operator equations of the first kind in partially ordered spaces are given. Section 1 presents some background information, relevant notions and theorems about regularizability. Section 2 describes certain techniques obtained by means of iterations (successive approximation) for solving variational inequalities with a monotone operator, as well as their applications to linear and quadratic programming. For solving unstable equations with non-monotonic operator the author considers regular one-and three-parameter iterative processes based on the Gauss-Newton method under some conditions on a test and required solutions. In Section 4 he presents another approach to regularization of the successive approximation method for problems with a nonexpanding or quasi-nonexpanding operator based on the introduction of correcting (damping) multipliers into computational schemes. Such an approach permits to ensure strong convergence for modifications of well-known methods (gradient projections method, prox-method), no matter whether the original problem is well-posed or not. Although methods of Gauss-Newton type allow to discard the non-monotonicity conditions, these impose rather rigid requirements upon the choice of a test solution and the a priori conditions for the required solution. In Section 5 by using partially ordered spaces techniques with isotone relative to a cone operator the author succeeds in establishing essentially weaker requirements on the initial approximations and in constructing iterative processes that monotonically converge to the solution. Finally, their applications to nonlinear integral equations are given.