question about variable bound
question about variable bound
Hi, my name is Eri. Right now I'm in my second year of master degree in chemical engineering. I'm using GAMS as a part in my thesis (big part actualy), since my work is very closely related with optimization. I have question about setting boundary.
My objective value (let say it's A) is about 0.8-0.9 (estimation), and I put safe upper bound (A.up), let say it's 5. Oddly, the result can be a different value (not 0.8/0.9/in between but a very small number). if I raise the upper bound to a bigger value (for example, 10), the result is normal (in 0.8-0.9 range).
The question is, is it good to have a very narrow range between lower and upper boundary, or is it normal to have odd result if the range is too narrow (hence, we have to have a wider range in order to give GAMS some kind of "space" to converge?).
my problem is minimization, and I am using baron solver (MINLP). Thanks in advance.
PS: I'm sorry for my English since it is not my first language ;D
--~--~---------~--~----~------------~-------~--~----~
To post to this group, send email to gamsworld@googlegroups.com
To unsubscribe from this group, send email to gamsworld+unsubscribe@googlegroups.com
For more options, visit this group at http://groups.google.com/group/gamsworld?hl=en
-~----------~----~----~----~------~----~------~--~---
Re: question about variable bound
Eri,
In most cases it is not a good idea to set any bounds (neither narrow
nor wide) on your objective variable. This prevents the solver to view
the objective constraint as a true objective function. Most solvers
work better if they get a true objective function.
Hope this helps.
Michael Bussieck - GAMSWorld Coordinator
On Oct 6, 4:38 am, "Eri Riyanto" wrote:
> > Hi, my name is Eri. Right now I'm in my second year of master degree in
> > chemical engineering. I'm using GAMS as a part in my thesis (big part
> > actualy), since my work is very closely related with optimization. I have
> > question about setting boundary.
> >
> > My objective value (let say it's A) is about 0.8-0.9 (estimation), and I put
> > safe upper bound (A.up), let say it's 5. Oddly, the result can be a
> > different value (not 0.8/0.9/in between but a very small number). if I raise
> > the upper bound to a bigger value (for example, 10), the result is normal
> > (in 0.8-0.9 range).
> >
> > The question is, is it good to have a very narrow range between lower and
> > upper boundary, or is it normal to have odd result if the range is too
> > narrow (hence, we have to have a wider range in order to give GAMS some kind
> > of "space" to converge?).
> >
> > my problem is minimization, and I am using baron solver (MINLP). Thanks in
> > advance.
> >
> > PS: I'm sorry for my English since it is not my first language ;D
--~--~---------~--~----~------------~-------~--~----~
To post to this group, send email to gamsworld@googlegroups.com
To unsubscribe from this group, send email to gamsworld+unsubscribe@googlegroups.com
For more options, visit this group at http://groups.google.com/group/gamsworld?hl=en
-~----------~----~----~----~------~----~------~--~---
Re: question about variable bound
Thank you so much for the information. I will change my code accordingly.
On Tue, Oct 7, 2008 at 10:30 AM, Gamsworld Admin wrote:
Eri,
In most cases it is not a good idea to set any bounds (neither narrow
nor wide) on your objective variable. This prevents the solver to view
the objective constraint as a true objective function. Most solvers
work better if they get a true objective function.
Hope this helps.
Michael Bussieck - GAMSWorld Coordinator
On Oct 6, 4:38 am, "Eri Riyanto" wrote:
> Hi, my name is Eri. Right now I'm in my second year of master degree in
> chemical engineering. I'm using GAMS as a part in my thesis (big part
> actualy), since my work is very closely related with optimization. I have
> question about setting boundary.
>
> My objective value (let say it's A) is about 0.8-0.9 (estimation), and I put
> safe upper bound (A.up), let say it's 5. Oddly, the result can be a
> different value (not 0.8/0.9/in between but a very small number). if I raise
> the upper bound to a bigger value (for example, 10), the result is normal
> (in 0.8-0.9 range).
>
> The question is, is it good to have a very narrow range between lower and
> upper boundary, or is it normal to have odd result if the range is too
> narrow (hence, we have to have a wider range in order to give GAMS some kind
> of "space" to converge?).
>
> my problem is minimization, and I am using baron solver (MINLP). Thanks in
> advance.
>
> PS: I'm sorry for my English since it is not my first language ;D
--~--~---------~--~----~------------~-------~--~----~
To post to this group, send email to gamsworld@googlegroups.com
To unsubscribe from this group, send email to gamsworld+unsubscribe@googlegroups.com
For more options, visit this group at http://groups.google.com/group/gamsworld?hl=en
-~----------~----~----~----~------~----~------~--~---
Re: question about variable bound
Still about setting boundary problem:
How to handle variables which lower and boundary I do not know and are not predictable because they do not have any physical meaning (hence, it is not possible to set a "safe" upper or lower boundary)? Right now I set them free, but then I counter feasibility problem (bounds too wide, model status uncertain - I'm using baron solver). Until now I have used brute force method only, and the program has to run for 2-3 hours before finding answers (some of them are not globally optimum). I have tried setting some arbitrary boundary (-1000 for lower bound and 1000 for upper bound) and this action, although increase the feasibility greatly, changed my objective true value.
Thank you
Eri
On Tue, Oct 7, 2008 at 2:06 PM, Eri Riyanto wrote:
Thank you so much for the information. I will change my code accordingly.
On Tue, Oct 7, 2008 at 10:30 AM, Gamsworld Admin wrote:
Eri,
In most cases it is not a good idea to set any bounds (neither narrow
nor wide) on your objective variable. This prevents the solver to view
the objective constraint as a true objective function. Most solvers
work better if they get a true objective function.
Hope this helps.
Michael Bussieck - GAMSWorld Coordinator
On Oct 6, 4:38 am, "Eri Riyanto" wrote:
> Hi, my name is Eri. Right now I'm in my second year of master degree in
> chemical engineering. I'm using GAMS as a part in my thesis (big part
> actualy), since my work is very closely related with optimization. I have
> question about setting boundary.
>
> My objective value (let say it's A) is about 0.8-0.9 (estimation), and I put
> safe upper bound (A.up), let say it's 5. Oddly, the result can be a
> different value (not 0.8/0.9/in between but a very small number). if I raise
> the upper bound to a bigger value (for example, 10), the result is normal
> (in 0.8-0.9 range).
>
> The question is, is it good to have a very narrow range between lower and
> upper boundary, or is it normal to have odd result if the range is too
> narrow (hence, we have to have a wider range in order to give GAMS some kind
> of "space" to converge?).
>
> my problem is minimization, and I am using baron solver (MINLP). Thanks in
> advance.
>
> PS: I'm sorry for my English since it is not my first language ;D
--~--~---------~--~----~------------~-------~--~----~
To post to this group, send email to gamsworld@googlegroups.com
To unsubscribe from this group, send email to gamsworld+unsubscribe@googlegroups.com
For more options, visit this group at http://groups.google.com/group/gamsworld?hl=en
-~----------~----~----~----~------~----~------~--~---