Publication: Computing Derivatives of Computer Programs
Introduction
Applications
Tools
Research Groups
Workshops
Publications
   List Publications
   Advanced Search
   Info
   Add Publications
My Account
About

Computing Derivatives of Computer Programs

- incollection -
 

Author(s)
C. H. Bischof , H. M. Bücker

Published in
Modern Methods and Algorithms of Quantum Chemistry: Proceedings, Second Edition

Editor(s)
J. Grotendorst

Year
2000

Publisher
NIC-Directors

Abstract
Automatic differentiation is introduced as a powerful technique to compute derivatives of functions given in the form of a computer program in a high-level programming language such as Fortran, C, or C++. In contrast to traditional approaches such as handcoding of analytic expressions, numerical approximation by divided differences, or manipulation of symbolic algebraic expressions by computer algebra systems, automatic differentiation offers the following substantial benefits: it is accurate up to machine precision, efficient in terms of computational cost, applicable to a 1-line formula as well as to a 100,000-line code, and can be produced with minimal human effort.

AD Theory and Techniques
General

BibTeX
@INCOLLECTION{
         Bischof2000CDo,
       author = "C.~H.~Bischof and H.~M.~B{\"u}cker",
       title = "Computing Derivatives of Computer Programs",
       booktitle = "Modern Methods and Algorithms of Quantum Chemistry: Proceedings, Second Edition",
       publisher = "NIC-Directors",
       editor = "J. Grotendorst",
       series = "NIC Series",
       pages = "315--327",
       address = "J{\"u}lich",
       url = "https://juser.fz-juelich.de/record/44658/files/Band_3_Winterschule.pdf",
       notes = "Also available as preprint ANL/MCS--P813--0400, Argonne National Laboratory,
         Mathematics and Computer Science Division, USA, April 2000",
       abstract = "Automatic differentiation is introduced as a powerful technique to compute
         derivatives of functions given in the form of a computer program in a high-level programming
         language such as Fortran, C, or C++. In contrast to traditional approaches such as handcoding of
         analytic expressions, numerical approximation by divided differences, or manipulation of symbolic
         algebraic expressions by computer algebra systems, automatic differentiation offers the following
         substantial benefits: it is accurate up to machine precision, efficient in terms of computational
         cost, applicable to a 1-line formula as well as to a 100,000-line code, and can be produced with
         minimal human effort.",
       ad_theotech = "General",
       year = "2000",
       volume = "3"
}


back
  

Contact:
autodiff.org
Username:
Password:
(lost password)