Nonlinear constraint

145 views
Skip to first unread message

ursula challita

unread,
Apr 21, 2013, 10:32:26 AM4/21/13
to ai...@googlegroups.com
Hello all,

I want to check whether the solvers in AIMMS solve non-linear problems or if 
anyone can suggest a way to linearize this constraint.

The nonlinear constraint is attached.

Thanks in advance for the help.
Ursula



Luis Pinto

unread,
Apr 21, 2013, 6:04:40 PM4/21/13
to AIMMS GoogleGroups
There are non-linear solvers in AIMMS (I don't work with non-linear, but I beleive Baron and Mosek are examples right?).

Not sure about the linearization... what are the variables?



--
You received this message because you are subscribed to the Google Groups "AIMMS - The Modeling System" group.
To unsubscribe from this group and stop receiving emails from it, send an email to aimms+un...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

ursula challita

unread,
Apr 22, 2013, 12:29:26 PM4/22/13
to ai...@googlegroups.com
Thanks Luis.

I've attached an updated form of my constraint.
I am looking for a solution approach, either using a non-linear solver or trying to linearize this objective.

I would appreciate any suggestion.
Thanks,
Ursula
Untitled.jpg

Vinay Sewdien

unread,
Apr 22, 2013, 1:43:33 PM4/22/13
to ai...@googlegroups.com
Hi Ursula,

The Identifier in which you have defined your mathematical problem has the option to chose the type of solver. I guess in your case it would be the nonlinear program NLP...

ursula challita

unread,
Apr 22, 2013, 3:13:43 PM4/22/13
to ai...@googlegroups.com
Thanks Vinay.
I wanted to check whether the solvers in AIMMS could solve such a problem.

Marcel Hunting

unread,
Apr 23, 2013, 8:23:50 AM4/23/13
to ai...@googlegroups.com
Hi,

I don't think there exists an equivalent linear formulation for your nonlinear constraint. Your problem appears to have binary variables and therefore the mathematical programming type becomes Mixed Integer Nonlinear Programming (MINLP) which is supported by the solvers AOA, BARON and Knitro in AIMMS. If your problem is non-convex (which it seems to be) then AOA and Knitro will only return a local optimum. BARON is a global solver but might take a long time to run. Unfortunately, BARON and Knitro are not available in the free academic license. An examples of running with AOA can be found in this AIMMS blog article: http://blog.aimms.com/2012/02/solving-minlp-problems-with-aimms-outer-approximation/ .

Marcel Hunting
AIMMS Software Developer

ursula challita

unread,
Apr 24, 2013, 7:36:57 AM4/24/13
to ai...@googlegroups.com
Hi Marcel,

Thanks a lot for the support.
Actually I was able to simplify the constraint further. The simplified version is attached.
I want to check also how can I get the license for Baron.

Many Thanks!

Ursula

Guido Diepen

unread,
Apr 24, 2013, 7:48:08 AM4/24/13
to ai...@googlegroups.com
Hi Ursula,

Baron is not included in the free academic license program of AIMMS.

If you want to have access to Baron, you will have to go for a paid academic license. For more details see our website: http://www.aimms.com/academic/paid-academic-license

Guido Diepen
AIMMS Specialist

Marcel Hunting

unread,
Apr 24, 2013, 9:43:31 AM4/24/13
to ai...@googlegroups.com
Hi Ursula,

What are the I_k and the I_j in the (1 - c_k I_k / I_j ) term? I assume that c_k and s_pi are the only variables in this constraint; is that correct?

Best regards,

Marcel

Ursula

unread,
Apr 24, 2013, 1:50:49 PM4/24/13
to ai...@googlegroups.com
Hello Marcel,

Indeed, c_i and s_pi are the binary variables in the problem.
I_k and I_j are parameters. 
Is there any approach that can linearize this objective?

Thanks for the help.
Ursula
You received this message because you are subscribed to a topic in the Google Groups "AIMMS - The Modeling System" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/aimms/-A0p4rr3q-E/unsubscribe?hl=en.
To unsubscribe from this group and all its topics, send an email to aimms+un...@googlegroups.com.

Marcel Hunting

unread,
Apr 25, 2013, 6:21:34 AM4/25/13
to ai...@googlegroups.com
Hi Ursula,

If I_k and I_j are 0/1 parameters then it would maybe be possible to linearize the constraint. If they are not then I see no way to linearize it.

Best regards,

Marcel

Ursula

unread,
Apr 25, 2013, 7:49:09 AM4/25/13
to ai...@googlegroups.com
Hi Marcel,

The parameters are not bounded. They take continuous values.
Maybe i'll try to solve it using a non-linear solver.

Thanks a lot for the support.
Ursula

ursula challita

unread,
Apr 28, 2013, 6:20:34 PM4/28/13
to ai...@googlegroups.com
Hello,

I want to check whether Baron is available in the free trial license?

Many Thanks!
Ursula
To unsubscribe from this group and all its topics, send an email to aimms+unsubscribe@googlegroups.com.

Guido Diepen

unread,
Apr 29, 2013, 6:22:55 AM4/29/13
to ai...@googlegroups.com
Hi Ursula,

Baron is indeed available in the free trial.

Guido Diepen
AIMMS Specialist
To unsubscribe from this group and all its topics, send an email to aimms+un...@googlegroups.com.

ursula challita

unread,
Apr 29, 2013, 7:44:21 AM4/29/13
to ai...@googlegroups.com
Hi Guido,

I installed a free trial license. Baron does not appear in the list of solvers.
How can I set it to be my default solver?

Many Thanks,
Ursula

Guido Diepen

unread,
Apr 29, 2013, 8:00:51 AM4/29/13
to ai...@googlegroups.com
Hi Ursula,

you  might have to install the Baron solver in the solver configuration screen. More information about how to install a new solver can be found in the Users Guide. You can find the details in the current version that is online  in Section 22.3 of http://www.aimms.com/aimms/download/manuals/aimms3ug_projectsettingsandoptions.pdf

Guido Diepen
AIMMS Specialist

ursula challita

unread,
Apr 30, 2013, 5:51:37 AM4/30/13
to ai...@googlegroups.com
Hi Guido,

Thanks a lot.
Ursula
Reply all
Reply to author
Forward
0 new messages