[Coin-ipopt] ensuring decrease of objective

[Andreas Waechter]

Hi James

What Graham describes is the intended way to give users the opportunity to 
stop the algorithm prematurely, or to collect some intermediate 
information.  So, if that is what you are interested in doing, that's the 
way to go.

But I'm not sure I understand you original question:  When you describe 
that the objective function increases at the end of the optimization, does 
that happen because you interrupt the optimization early?  (I assume here 
that you have no constraints except for bound constraints for the 
variables...)  In that case you might want to try if the optimization is 
better when you either use a small initial value of the barrier parameter

mu_init 1e-4   [or some other small value]

or if you use the automatic choice of the barrier parameter

mu_strategy adaptive


Regards,

Andreas


On Mon, 23 Apr 2007 g_sparrow at hotmail.com wrote:

> One thought on this... If you are using the C++ interface, you can implement 
> the intermediate_callback() virtual function. By doing this, you can 
> implement whatever heuristics you want to cause the algorithm to terminate. 
> For example you could set the IPOPT max_iter quantity to some large value and 
> then implement your own termination criteria which uses a combination of the 
> number of iterations, with the additional check that the objective must have 
> decreased. Of course, stopping the algorithm prematurely means that you are 
> bypassing IPOPT's own convergence checking, so you may need to some 
> additional checks as well (e.g. feasibility etc.).
>
> There may be other (better) ways of doing this, but this is very flexible.
>
> Graham
>
> ----- Original Message -----
> From: "James Lu" <james.lu at oeaw.ac.at>
> To: <coin-ipopt at list.coin-or.org>
> Sent: Monday, April 23, 2007 4:27 AM
> Subject: [Coin-ipopt] ensuring decrease of objective
>
>> hi:
>> 
>> We are using IPOPT as a local solver within scatter search, for parameter
>> identification in gene networks. For some of the scatter search heuristics, 
>> it
>> is necessary that the objective decreases after carrying out the local 
>> solve.
>> However, due to the box constraints and the small number of IPOPT 
>> iterations we
>> carry out, sometimes the objective actually increases at the end of the 
>> local
>> optimization. Is there a way to set the options so that one carries out 
>> enough
>> iteations to at least ensure the objective DECREASES from is initial value?
>> 
>> thank you,
>> 
>> James
>> 
>> -- 
>> Dr James Lu
>> 
>> Research Scientist
>> Johann Radon Institute for Computational and Applied Mathematics
>> 
>> _______________________________________________
>> Coin-ipopt mailing list
>> Coin-ipopt at list.coin-or.org
>> http://list.coin-or.org/mailman/listinfo/coin-ipopt
>> 
> _______________________________________________
> Coin-ipopt mailing list
> Coin-ipopt at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/coin-ipopt
>