All well-posed problems have uniformly stable and convergent discretizations (Q5964947)

This paper considers the discretization of a large class of well-posed linear operator equations, which include linear boundary value problems. The starting point is to formulate the problem as linear recovery problems for functions from the data, where the well-posedness means the continuity of the recovery. The discretization recovers restricted trial functions from restricted test data, and likewise, the well-posedness means the restricted recovery is continuous. The paper formulates an abstract framework for the discretization using the concept of monotonic refinable dense (discretization), and shows that all well-posed linear problems have stable and refinable computational discretizations with a stability that is determined by the well-posedness of the continuous problem only, provided that sufficiently many test data are used. However, the proof is not constructive, and does lead directly to a bound on the number of test data needed. An abstract convergence rate in terms of the approximation error in the data space is also given. The framework covers many interesting cases, e.g., generalized finite elements, symmetric and unsymmetric Kansa-type collocation, and meshless local Petrov-Galerkin method. The case of noisy data is also briefly discussed.