[Ipopt] infeasible solution?

[Chunhua Men]

Hi Stefan, I just sent an attachment of the screenshot to you, but then I
was told:


Your mail to 'Ipopt' with the subject

    Re: [Ipopt] infeasible solution?

Is being held until the list moderator can review it for approval.

The reason it is being held:

    Message body is too big: 92768 bytes with a limit of 40 KB

On Thu, Jun 23, 2016 at 10:02 AM, Stefan Vigerske <stefan at math.hu-berlin.de>
wrote:

> Hi,
>
> can you send the Ipopt log?
> It's hard to believe that the constraint violation is exactly 0.
>
> I would think that it reports convergence to a point of local
> infeasibility, if it gets to a point in which the constraint violation is
> above the feasibility tolerance ("tol", in the simplest case) and it
> doesn't see a direction in which it can reduce the infeasibility.
> For a feasible convex NLP, this should not happen usually.
>
> Stefan
>
>
> On 06/23/2016 05:37 PM, Chunhua Men wrote:
>
>> Hi, I used IPOPT to solve a convex NLP model. The starting point
>> (initial solution) is feasible, but after <100 iterations, IPOPT always
>> reports:
>>
>> "EXIT: Converged to a point of local infeasibility. Problem may be
>> infeasible”.
>>
>> I printed out the final solution: each decision variable is within the
>> bounds, and all constraints are met.  From the black console, IPOPT also
>> reports "Constraint violation" is zero.  So, it is a feasible solution
>> (though
>> may not be optimal).
>>
>> May I know in which situation, IPOPT would report "Converged to a point of
>> local infeasibility"?
>>
>> My IPOPT version is 3.11.9.9 and it was built in VS 2012 (64-bit) - Win7.
>>
>> Thanks! Chunhua
>>
>>
>>
>> _______________________________________________
>> Ipopt mailing list
>> Ipopt at list.coin-or.org
>> http://list.coin-or.org/mailman/listinfo/ipopt
>>
>>
>
> --
> http://www.gams.com/~stefan
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/ipopt/attachments/20160623/f0b93c48/attachment.html>