Automatic differentiation of iterative processes (Q1334755)

Automatic differentiation is a method for computing the derivative values of a function \(f: D\subseteq \mathbb{R}^ n\to \mathbb{R}\), which is given in the form of an algorithm A. By using automatic differentiation it is intended to convert A into another algorithm \(\text{A}'\), which describes the forward mode of automatic differentiation. It permits to calculate the values of the derivative of a real-valued function given by a code list. In this paper the differentiation of a function \(f\) defined by an infinite code list or iterative process is studied. From previous investigations of application of automatic differentiation to iterative processes the contributions of \textit{J. C. Gilbert} [Optim. Methods Software 1, 13-21 (1992)] should be mentioned. The author ends his paper stating that for some special iterative processes, like some Newton methods, it can be proved that the differentiation can be done automatically.