ADOL-C: Computing higher-order derivatives and sparsity pattern for functions written in C/C++

Abstract:

This paper presents ADOL-C, a software package for the Automatic Differentiation of C and C++ codes. Based on the operator overloading facility in C++, ADOL-C generates as first step an internal representation of the function to be differentiated. Subsequently, derivative information of arbitrary order can be computed using either the forward or the reverse mode. Furthermore, the chosen implementation strategy allows the calculation of structural information, e.g. sparsity patterns of Jacobians. After discussing the main ideas of this approach to implement AD, we will present applications of ADOL-C for optimization problems from mechanical engineering as well as chemical engineering. However, AD may be useful even for pure simulation as illustrated by another example, namely the integration of higher order DAEs.

Año de publicación:

2004

Keywords:

• Sparsity detection
• Forward mode
• Reverse mode
• Automatic differentiation
• Higher-order derivatives

Fuente:

scopus