[Coin-ipopt] AMPL

[Andreas Waechter]

Hi Susi,

The advantiages of using Ipopt thorugh AMPL is that

- you can easily experiment with different model formulations and see
   which is best for your problem
- you don't need to worry about writing C++ code for the derivatives of
   the functions - this is very often a headache because one easily makes
   mistakes, while AMPL does that for you

A disadvantage of AMPL is that you need to have a license for AMPL (unless 
your problem has only less than 300 variables and constraints, in which 
case the student AMPL edition works).  But you can use AMPL remotely at 
NEOS (http://www-neos.mcs.anl.gov/neos/solvers/nco:Ipopt/AMPL.html) - I 
have to admit that we still didn't update the Ipopt verison available 
there, it is still the 2.2.1 Fortran version...

Also, if you have a large problem (say, more than 10000 variables), AMPL 
can be come somewhat inefficient, and then it will be faster to implement 
the problem statement in C++.

I usually suggest to users to use AMPL to experiment with the model 
formulation for their application, and see how Ipopt (or other solvers) 
perform.  If then AMPL for the real application because of speed or 
licensing issues, I would then write C++ code to implement the problem 
formulation that I found to work best in the AMPL experiments.

I hope this helps,

Andreas

On Thu, 20 Dec 2007, Susana Lera wrote:

> Hi,
>
> i'm using IPOPT to fins the optimal solution for a low thrust orbit transfer. Since now i've been using the c++ interface, but i can't get a proper optimal solution.
> could it be better for my case to use the AMPL interface?
> Thanks.
>
> susi.
>
>
> _______________________________________________
> Coin-ipopt mailing list
> Coin-ipopt at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/coin-ipopt
>