[Ipopt] Can i use IPOPT without the input of Jacobian Matrix and the Hessian of the Lagrangian?

[Uwe Nowak]

Also I am not member of the IPOPT-Team:

You do not need the Hessian if you use quasi-newtown approximation of 
second derivatives (LBFGS). Set the option
hessian_approximation=limited-memory

However you need the Jacobian.
If you have a set of constraints and are not sure which constraints are 
used, you could define the Jacobian structure as being all possible 
entries of any element.

If you do not have the Jacobian functions you either have to do finite 
difference approximation (which might lead to slow evaluation and 
convergence problems) or automatic differentiation tool (I have heard 
about ADOL-C but never used it).

Best Regards,
Uwe


Am 16.12.2009 10:54, schrieb Lewis I:
> Hello Ipopt-Team,
>
> I am a MPhil. Student at the Simon Fraser University. I am working on a research which requires the nonlinear optimization.
>
> I would like to use Ipopt. I have tried the pre-complied binaries for Windows (C++) and read the examples.
> And I found that Ipopt require the input of Jacobian Matrix and the Hessian of the Lagrangian before the calculaion of the optimization.
> Because one of the requirement of my program is to allow it can be more interactive.
> I can not predefine the Jacobian Matrix and the Hessian of the Lagrangian for all constraints, because the constraints will be changed at different situations.
> Do there exist any method to compute these matrix from the set of provided constraint functions automatically?
>
> Thank you very much for your help.
>
> Best regards
> Lewis
>
>
>
>
>
>
>
> _______________________________________________
> Ipopt mailing list
> Ipopt at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/ipopt