[Ipopt] MATLAB interface for IPOPT and AMPL models

Gianpaolo Gulletta gianpaologulletta at gmail.com
Thu Feb 19 07:07:42 EST 2015


Hi Tony,

thank you for your quick reply.

I think I will move to Jump and Julia because I need high performance. I 
hope that translating my model in ampl language (.mod files) into 
"julia-like" models will not be so hard. My main problem is that I 
cannot calculate exact gradient functions of my constraints because the 
problem is an highly non-linear large-scale problem.

Thank you.
Regards,
Gianpaolo
On 19-02-2015 11:52, Tony Kelman wrote:
> No. The Matlab interface requires you to implement objective and 
> constraint functions, gradient and Jacobian functions, and optionally 
> the Hessian of the Lagrangian, all as Matlab functions. You can use a 
> modeling environment like Yalmip (or CVX maybe? I don't think that has 
> an interface to Ipopt hooked up right now, but it could be possible in 
> theory) to translate the problem from AMPL syntax into Matlab syntax, 
> and it will automatically calculate the derivatives for you. JuMP in 
> Julia is similar, and much higher performance.
>
> The command-line Ipopt executable can read .nl files, but you need a 
> licensed version of AMPL to create a .nl from .mod and .dat files.
>
> -Tony
>
>
> -----Original Message----- From: Gianpaolo Gulletta
> Sent: Thursday, February 19, 2015 2:07 AM
> To: ipopt at list.coin-or.org
> Subject: [Ipopt] MATLAB interface for IPOPT and AMPL models
>
> Dear all,
>
> I would like to know if there is a way to use the matlab interface for
> ipopt in conjunction with an AMPL model.
>
> Basically, I have one AMPL model file (.mod) and three data files (.dat)
> all written in AMPL language. I want to read them all and solve the 
> problem.
>
> Is it possible by mean of the MATLAB interface?
>
> Thank you.
>
> Best Regards,
>

-- 
Gianpaolo Gulletta



More information about the Ipopt mailing list