On optimal direct methods for solving Volterra equations in a Hilbert space (Q1316925)

By a direct method for the approximate solution of an operator equation of the second kind \(z = Hz + f\) the authors mean a rule \(D\) which to each operator \(H\) of a given class \({\mathcal H}\) associates a coordinate system of elements \(\varphi_ k = \varphi_ k(D)\); \(k=1, \dots,N\); and to each pair \((H,f)\) a collection of coefficients \(a_ k=a_ k (H,f,D)\); \(k = 1, \dots, N\); such that \(z_ N = z_ N (H,f,D) = \sum^ N_{k=1} a_ k \varphi_ k + \varphi_ 0\). The problem is to find a ``best`` method \(D\) for fixed \(N\). The error of the Galerkin iterative method is estimated. A main result of the paper is: The optimal order of certain classes of Volterra equations defined in this paper is realized by the Galerkin iterative method. In some classes of Volterra equations the error of optimal nonadaptive direct methods is worse than the error of the optimal adaptive direct methods. There are some remarks about the possibility of the effective numerical implementation of the Galerkin iterative method. Unfortunately, a lot of indices is not printed correctly in the English translation. By the way, inequality (10) is not a consequence but an essential part of the definition of a class of integral operators given around formula (9).