[Ipopt] Setting initial feasible point

Andreas Waechter andreasw at watson.ibm.com
Mon Apr 11 09:06:31 EDT 2011


Hi Akshay,

Yes, Ipopt changes the starting point so that it is sufficiently in the 
interior of the region defined by the bounds.  The details are described 
in Section 3.6. in the implementation paper (see Ipopt home page), and the 
corresponding Ipopt options are bound_push and bound_frac.  If you want 
less perturbation of the starting point, you generally choose smaller 
values of these parameters.  And if you indeed have a good starting point 
close to the solution, you might also want to try to reduce mu_init (or 
try mu_strategy=adaptive) and maybe play with the the initialization of 
the bound multipliers (e.g., bound_mult_init_method=mu-based).

Good luck,

Andreas

On Sat, 9 Apr 2011, Akshay Gupte wrote:

> Hi all,
> I am using the AMPL interface for solving my NLPs using Ipopt. I would like to set a starting point for Ipopt. How do I go about it ? I tried the usual AMPL syntax :
> let {i in I} x[i] := startval[i];
>
> Although, this works with other NLP solvers, Ipopt does not seem to recognize this initial point. I read that initial feasible points are handled differently by interior point solvers to ensure that they are not too close to the boundary. Do I need to set any warm starting parameters, bound_frac etc. ?
>
> Any help is much appreciated.
>
> Regards,
> Akshay Gupte
> PhD student, Operations Research
> Georgia Institute of Technology
> akshayg at gatech.edu
>
>
>
>
>
>
>
> _______________________________________________
> Ipopt mailing list
> Ipopt at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/ipopt
>
>



More information about the Ipopt mailing list