A Hitchhiker's guide to automatic differentiation (Q2630759)

The focus of this article is to provide short overviews of the distinct descriptions of automatic differentiation (AD) and to show why all of these expressions are of the same principle. This method is of importance since it does not have the well-known problems of approximate methods (truncation error) or symbolic differentiation (significantly long computation times). Then only error of relevance is that incurred due to rounding errors in floating-point arithmetic or due to imprecise evaluations of elementary functions. The focus is on one distinct version (or mode) of AD, namely the forward mode. The purpose of the text is predominantly educational.