An ADA library for automatic evaluation of derivatives (Q751170)

The paper describes in detail an implementation in ADA language of a set of simple procedures for evaluating derivatives exactly. This evaluation requires no more than the writing of the original function in the usual way using standard mathematical operators and intrinsic functions. The method is applied to nonlinear least squares problems. The ordinary Gauss-Newton algorithm is used but only for problems with two independent variables and two unknown parameters.Two practical numerical examples of parameter estimation are given and a listing of the ADA library used for the evaluations in this paper is included.