Automatic Differentiation for Iterative Process and Its Applications in Network Performance Analysis

Abstract

In this paper we focus on the application of automatic differentiation (AD)technique on iterative processes. We review some of the results on theconvergence of general iterative processes and the convergence of thederivative code of such iterative processes. We are especially interestedin a class of fixed point iteration problems and we extended some of thoseresults to discuss this class of problems. Finally we apply an AD packageADIC to a network performance evaluation problem for numerical experimentsto get sensitivities of network blocking probabilities w.r.t. networkoffered traffic load.