MIP Infeasibility

GAMS
1

Hi All,
I just have a very simple cost minimization MIP but seem to run into some feasibility issues for no reason as the same construct/code works fine with any other modeling platform like AIMMS.

demand(d,t)…sum(r,nc(r,d)*X(r))=g=sysdemand(d,t);
r={1,2,3}
d={1,2,3}
nc(r,d) = Integer Variable
sysdemand(d,t)= Demand at some destination nodes. modeled as a table
X(r)= Parameter

This constraint is always infeasibile as the integer variable is set to zero . But when I change the RHS to some fixed number say 500,600, the code works fine. I import the same code and try it in AIMMS, I do not have any in feasibility issues at all. Why is GAMS defaulting the nc(,) variable to zero all the time?.

–
To unsubscribe from this group and stop receiving emails from it, send an email to gamsworld+unsubscribe@googlegroups.com.
To post to this group, send email to gamsworld@googlegroups.com.
Visit this group at http://groups.google.com/group/gamsworld.
For more options, visit https://groups.google.com/groups/opt_out.

\

2

Just a quick thought. GAMS has an upper limit on integer variables set to 100 by default (don’t ask me why). That might cause the issues. Update to this in the model

IntegerVariable.up = inf; (or something sufficiently high)

(or you can set an execution option parameter pf4=0 I think).


On Mon, Jun 24, 2013 at 9:40 AM, Ragu wrote:

Hi All,
I just have a very simple cost minimization MIP but seem to run into some feasibility issues for no reason as the same construct/code works fine with any other modeling platform like AIMMS.

demand(d,t)…sum(r,nc(r,d)*X(r))=g=sysdemand(d,t);
r={1,2,3}
d={1,2,3}
nc(r,d) = Integer Variable
sysdemand(d,t)= Demand at some destination nodes. modeled as a table
X(r)= Parameter

This constraint is always infeasibile as the integer variable is set to zero . But when I change the RHS to some fixed number say 500,600, the code works fine. I import the same code and try it in AIMMS, I do not have any in feasibility issues at all. Why is GAMS defaulting the nc(,) variable to zero all the time?.

–
To unsubscribe from this group and stop receiving emails from it, send an email to gamsworld+unsubscribe@googlegroups.com.
To post to this group, send email to gamsworld@googlegroups.com.
Visit this group at http://groups.google.com/group/gamsworld.
For more options, visit https://groups.google.com/groups/opt_out.



\

To unsubscribe from this group and stop receiving emails from it, send an email to gamsworld+unsubscribe@googlegroups.com.
To post to this group, send email to gamsworld@googlegroups.com.
Visit this group at http://groups.google.com/group/gamsworld.
For more options, visit https://groups.google.com/groups/opt_out.

\

3

Hi Cooper
Thanks for the reply. I did try that and the infeasibility seems to have vanished. The problem I have taken is very straight forward and simple but the absolute gap is high. The solution in AIMMS and GAMS do not seem to match at all. Wonder why such issues and trying hard to understand which is the right solution from solving/math point of view rather than the best answer.

On Monday, June 24, 2013 10:40:49 AM UTC-4, Ragu wrote:

Hi All,
I just have a very simple cost minimization MIP but seem to run into some feasibility issues for no reason as the same construct/code works fine with any other modeling platform like AIMMS.

demand(d,t)…sum(r,nc(r,d)*X(r))=g=sysdemand(d,t);
r={1,2,3}
d={1,2,3}
nc(r,d) = Integer Variable
sysdemand(d,t)= Demand at some destination nodes. modeled as a table
X(r)= Parameter

This constraint is always infeasibile as the integer variable is set to zero . But when I change the RHS to some fixed number say 500,600, the code works fine. I import the same code and try it in AIMMS, I do not have any in feasibility issues at all. Why is GAMS defaulting the nc(,) variable to zero all the time?.

–
To unsubscribe from this group and stop receiving emails from it, send an email to gamsworld+unsubscribe@googlegroups.com.
To post to this group, send email to gamsworld@googlegroups.com.
Visit this group at http://groups.google.com/group/gamsworld.
For more options, visit https://groups.google.com/groups/opt_out.

\

4

Hi: You can overcome the 0 defaulting by setting your own start values with nc.l (i e: nc(‘1’,‘2’).l = 1;).

Regards
Claudio



On Mon, Jun 24, 2013 at 11:51 AM, Ryan Cooper wrote:

Just a quick thought. GAMS has an upper limit on integer variables set to 100 by default (don’t ask me why). That might cause the issues. Update to this in the model

IntegerVariable.up = inf; (or something sufficiently high)

(or you can set an execution option parameter pf4=0 I think).


On Mon, Jun 24, 2013 at 9:40 AM, Ragu wrote:

Hi All,
I just have a very simple cost minimization MIP but seem to run into some feasibility issues for no reason as the same construct/code works fine with any other modeling platform like AIMMS.

demand(d,t)…sum(r,nc(r,d)*X(r))=g=sysdemand(d,t);
r={1,2,3}
d={1,2,3}
nc(r,d) = Integer Variable
sysdemand(d,t)= Demand at some destination nodes. modeled as a table
X(r)= Parameter

This constraint is always infeasibile as the integer variable is set to zero . But when I change the RHS to some fixed number say 500,600, the code works fine. I import the same code and try it in AIMMS, I do not have any in feasibility issues at all. Why is GAMS defaulting the nc(,) variable to zero all the time?.

–
To unsubscribe from this group and stop receiving emails from it, send an email to gamsworld+unsubscribe@googlegroups.com.
To post to this group, send email to gamsworld@googlegroups.com.
Visit this group at http://groups.google.com/group/gamsworld.
For more options, visit https://groups.google.com/groups/opt_out.



\

To unsubscribe from this group and stop receiving emails from it, send an email to gamsworld+unsubscribe@googlegroups.com.
To post to this group, send email to gamsworld@googlegroups.com.
Visit this group at http://groups.google.com/group/gamsworld.
For more options, visit https://groups.google.com/groups/opt_out.



\

To unsubscribe from this group and stop receiving emails from it, send an email to gamsworld+unsubscribe@googlegroups.com.
To post to this group, send email to gamsworld@googlegroups.com.
Visit this group at http://groups.google.com/group/gamsworld.
For more options, visit https://groups.google.com/groups/opt_out.

\

5

Oh you should probably set your optimality gap as well if you haven’t
put your modelname in the pace below before the solve statement

(modelname).optcr = .0005;


On Mon, Jun 24, 2013 at 10:05 AM, Ragu wrote:

Hi Cooper
Thanks for the reply. I did try that and the infeasibility seems to have vanished. The problem I have taken is very straight forward and simple but the absolute gap is high. The solution in AIMMS and GAMS do not seem to match at all. Wonder why such issues and trying hard to understand which is the right solution from solving/math point of view rather than the best answer.


On Monday, June 24, 2013 10:40:49 AM UTC-4, Ragu wrote:

Hi All,
I just have a very simple cost minimization MIP but seem to run into some feasibility issues for no reason as the same construct/code works fine with any other modeling platform like AIMMS.

demand(d,t)…sum(r,nc(r,d)*X(r))=g=sysdemand(d,t);
r={1,2,3}
d={1,2,3}
nc(r,d) = Integer Variable
sysdemand(d,t)= Demand at some destination nodes. modeled as a table
X(r)= Parameter

This constraint is always infeasibile as the integer variable is set to zero . But when I change the RHS to some fixed number say 500,600, the code works fine. I import the same code and try it in AIMMS, I do not have any in feasibility issues at all. Why is GAMS defaulting the nc(,) variable to zero all the time?.

–
To unsubscribe from this group and stop receiving emails from it, send an email to gamsworld+unsubscribe@googlegroups.com.
To post to this group, send email to gamsworld@googlegroups.com.
Visit this group at http://groups.google.com/group/gamsworld.
For more options, visit https://groups.google.com/groups/opt_out.



\

To unsubscribe from this group and stop receiving emails from it, send an email to gamsworld+unsubscribe@googlegroups.com.
To post to this group, send email to gamsworld@googlegroups.com.
Visit this group at http://groups.google.com/group/gamsworld.
For more options, visit https://groups.google.com/groups/opt_out.

\