[Ipopt] feasibility refinement mod

[Martin Neuenhofen]

I forgot to say that I already do so :)

Thank you anyway.

2017-01-19 14:52 GMT+00:00 Stefan Vigerske <stefan at math.hu-berlin.de>:

> Maybe enabling start_with_resto could help:
> https://www.coin-or.org/Ipopt/documentation/node50.html#SECT
> ION0001110040000000000000
>
> On 01/19/2017 12:48 PM, Martin Neuenhofen wrote:
>
>> Dear community,
>>
>> I observe that the solver does not try to achieve feasibility as long as
>> an
>> improvement in the cost function can be made. However, at some point I
>> want
>> so say "I am happy without optimality in the 10th digit of my cost
>> function, so now stop and make it feasible.".
>> So what I tried first of all was the option
>> options.ipopt.acceptable_obj_change_tol = 1 and hoped the solver would
>> directly from the first iter only go for feasibility in order to stop as
>> soon as possible. However, it still solves until a few dozens of
>> iterations
>> and only eventually decreases the primal infeasibility.
>>
>> So my questions are:
>>
>>    - May it be that the option acceptable_obj_change_tol does not have any
>>    effect at all in Matlab Ipopt 3.12.3 ?
>>    - How can I force Ipopt into only finding a feasible point from my
>>    current guess as quickly as possible without increasing the current
>> cost
>>    function value? So I want to do early what Ipopt would do in the few
>> very
>>    last iterations where it refines the feasibility.
>>
>> Cheers,
>> Martin
>>
>>
>>
>> _______________________________________________
>> 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/20170119/5408da9a/attachment-0001.html>