We propose a new class of projection methods for the solution of ill-posed problems with inaccurately specified coefficients. For methods from this class, we establish the conditions of convergence to the normal solution of an operator equation of the first kind. We also present additional conditions for these methods guaranteeing the convergence with a given rate to normal solutions from a certain set.
Statement of the ProblemIn recent years, the investigations of the problems of construction of economic finite-dimensional approximations to the solutions of ill-posed problems with given rates of convergence are extensively developed (see, e.g., [1] and references therein). It is known that the methods based on the use of Galerkin's projection scheme are the most widespread among the algorithms of approximate solution of the indicated problems. It turns out that the modification of Galerkin's method constructed by using step hyperbolic crosses is especially economic. In other words, this modification enables one to construct approximation with required accuracy by using much smaller amounts of discrete information than the standard Galerkin method. The first result of efficient application of the indicated modified projection scheme to the solution of ill-posed problems was obtained in [2]. This work was continued in numerous publications, among which we especially note [3 -5]. In these papers, new projection methods for the solution of ill-posed problems were proposed on the basis of the idea of hyperbolic cross. These methods guarantee the possibility of approximation of the solution with given rate of convergence under the assumption that the required solution belongs to a certain compact set. At the same time, an important aspect of substantiation of any approximate method, is connected with the analysis of conditions of its convergence to the exact solution of the original problem without any additional assumptions about this solution. In this sense, the present paper can be regarded as a continuation and generalization of [5]. Indeed, within the framework of investigations performed in what follows, we generalize the approach to the solution of ill-posed problems proposed in [5] to the case where the coefficients of the original equation are given with certain errors and establish sufficient conditions for the convergence of the constructed approximations to the required solution.Consider an operator equation of the first kindin a separable Hilbert space ( , ) ⋅ ⋅ with scalar product ( , ) ⋅ ⋅ and norm x = ( , ) x x . Suppose that A ∈ L ( ) X , Rang( ) A ≠ Rang( ) A , and f A ∈Rang( ), where L ( ) X is a space of linear continuous operators acting in X. The norm in L ( ) X is defined in the standard way: