🚀 go-pugleaf

ajayshah@almaak.usc.edu (Ajay Shah) wrote (sorry for missing details of the
original posting, it has vanished from our system):

> It looks like the best way to do automatic differentiation is to use
> C++.  What follows gives you some pointers to software available.

You might be right, although I suspect that the overheads of a C++
implementation could be significant.  It is funny, but in the literature
there are amazingly few efficiency evaluations on real-life problems (e.g.
compared against numerical differences).

>> Title: Automatic Differentiation of Algorithms: Theory, Implementation,
>> 	and Application
>> Edited by	Andreas Griewank, Argonne National Laboratory
>> 		George F. Corliss, Marquette University
>> Publisher: SIAM
>> ISBN 0-89871-284-X
>>
>> Proceedings of the first SIAM Workshop on Automatic Differentian, held
>> in Breckenridge, Colorado, January 6-8, 1991.

Yes, this is probably the best reference to date.  This book includes also
a review by Juedes of most of the AD software available.

> Here is a bibliography on automatic differentiation, from my (
> i.e., ANS) .bib:
>
> @techreport{Garcia91,
>	author = 	{Oscar Garcia},
>	title = 	{Experience with automatic differentiation in estimating forest growth model parameters},
>	institution = 	{Forest Research Institute},
>	month = 	{March},
>	year = 		1991,
>	address = 	{Rotorua, New Zealand},
> }

This is one of two drafts that I handed out at the Breckenridge Workshop. 
The other one described my GRAD AD system.  Both were combined in the paper
published in the Proceedings mentioned above.

GRAD is freely available, and I could email it, although I would prefer to
send it to some FTP site.  However, note that: a) It differentiates Fortran
code, producing a stand-alone Fortran routine. b) It is written in APL,
although knowledge of APL is not required and there is a (very slow)
version for the free I-APL interpreter.  I won't get into religious
arguments about Fortran vs C++ or whatever.  I needed it for an application
that was already in Fortran, and also the Fortran syntax was a little
easier to handle.  Besides, APL is obviously superior to anything else :-)

I am moving, and I may lose this e-mail address at any time.  New address
from mid-September 1992:  

Dr Oscar Garcia
Instituto Forestal
Huerfanos 554
Santiago - CHILE    (don't know about e-mail yet)