Re: Automatic Differentiation in Mathematica

*To*: mathgroup at smc.vnet.net*Subject*: [mg85591] Re: Automatic Differentiation in Mathematica*From*: Jens-Peer Kuska <kuska at informatik.uni-leipzig.de>*Date*: Thu, 14 Feb 2008 06:26:13 -0500 (EST)*Organization*: Uni Leipzig*References*: <foumc1$qv0$1@smc.vnet.net> <fp0kus$5cs$1@smc.vnet.net>*Reply-to*: kuska at informatik.uni-leipzig.de

Hi, suppose you have a function of several hundred or thousend variables but less that a million. And you want to minimize it. You have written a FORTRAN or C function that compute this function with everything you can do in FORTRAN, i.e., DO-loops, IF ... To minimize the function you need the gradient, but the computation of the gradient numerical is to error sensitive and take too long. Automatic Differntation take you FORTRAN or C function and generate a FORTRAN or C function that compute the gradient. And yes it compute the derivative of a DO-loop and nested IF statements. AFAIK there is no Mathematica package that does this. Regards Jens Daniel Huber wrote: > Hi marco, > can you please explain what you mean by: > "Differentiation of an algorithm" > Usually we differentiate functions. For symbolic differentiation there > is D and Dt. For numeric differentiation there is ND > sincerely, Daniel > > Marco Gabiccini wrote: >> Hi, >> >> I was wondering if any of you know about packages in Mathematica implementing >> Automatic Differentiation of algorithms like ADOL-C in C++ or MAD in another system. >> >> Thanks, >> >> marco >> >> Ing. Marco Gabiccini, PhD >> Assistant Professor of Applied Mechanics >> Dept. of Mechanical, Nuclear and Production Engineering >> University of Pisa >> Via Diotisalvi, 2, 56126 Pisa (PI) - Italy >> Phone: +39-050-836.676 >> E-mail: m.gabiccini at ing.unipi.it >> >> >