[Ipopt] Restoration phase is called at point that is almost feasible, with constraint violation 4.202374e-007

[Hans Pirnay]

Hi everyone,

I looked at the revisions and found that the change mentioned by
Andreas in his release announcement for 3.8.1 is in revision 1596 in
file Ipopt/src/Algorithm/IpIpoptData.cpp. It seems to me that you want
to go back to pre-3.8.1-restoration settings, you just need to change
line 46 in IpIpoptData.cpp from

#if 0

to

#if 1

and recompile. That way, the tolerance for the restoration phase is
set to 1e-2 times the tolerance for the regular algorithm.

cheers

Hans

On Tue, Jan 26, 2010 at 12:01 PM, James Henderson
<jamesh125125 at yahoo.com> wrote:
> Hi Ali and everyone,
>
> Do anyone know which "  termination tolerance" option should be changed for solving the restoration phase problem?
> Thank you.
>
> James
>
>
>
> ----- Original Message ----
> From: Ali Baharev <ali.baharev at gmail.com>
> To: Sam Hall <samh_1205 at yahoo.co.uk>
> Cc: ipopt at list.coin-or.org
> Sent: Tue, January 26, 2010 6:27:47 PM
> Subject: Re: [Ipopt] Restoration phase is called at point that is almost feasible, with constraint violation 4.202374e-007
>
> Hello Everyone,
>
> Please let me start with declaring that I am a great fan of the
> restoration phase in IPOPT.
>
> I am writing a research paper demonstrating how powerful and robust it
> is: it can solve problems that are difficult or even (practically)
> impossible to solve with other state-of-the-art commercial solvers!
>
> I do not have the time to have a look at each of these problems but
> they may have a common cause:
>
> http://list.coin-or.org/pipermail/ipopt/2009-July/001646.html
>
> In the announcement
>
> http://list.coin-or.org/pipermail/ipopt/2009-November/001732.html
>
> the above issue is said to be solved by "loosening the termination
> tolerance for the restoration phase problem" but it seems to me a
> workaround. Why did a strict termination work just fine in 3.6.1?
>
> http://list.coin-or.org/pipermail/ipopt/2010-January/001845.html
>
> http://list.coin-or.org/pipermail/ipopt/2010-January/001847.html
>
> http://list.coin-or.org/pipermail/ipopt/2010-January/001853.html
>
> Developers, please check if it was not a change in the restoration
> phase that causes the above issues.
>
> I will test the new version of IPOPT on my benchmark set in the next
> couple of weeks and see that the restoration phase is still as robust
> as it was in 3.6.1.
>
> Thanks,
>
> Ali
>
> _______________________________________________
> Ipopt mailing list
> Ipopt at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/ipopt
>
>
>
>
>
> _______________________________________________
> Ipopt mailing list
> Ipopt at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/ipopt
>