Abstract
In this paper the authors consider the problem of Laplace transform inversion in the case where the transform is given in a finite set of equidistant points. They propose a criterion, based on cross-validation, for the optimal choice of the first data point.