Variable enters more than one =C= equation

GAMS
1

Hi all,

Sorry for the disturbance.

I am using the GAMS to solve the second order cone programming (SOCP). I model it by using =c= to specify the conic constraints and then I can use CPLEX to solve the resulting “LP” problem. I have several conic constraints and they share the same decision variables: x. The model seems correct but the GAMS shows that “Variable enters more than one =C= equation”. Does anyone know what’s going on here? Can I still use =c= equations (LP model) and solve the problem?

If I use the common NLP model and solve it by using CONOPT. Is there any difference in terms of the solving time or result compared with the “LP” model solved by CPLEX. This is convex so I think the result will be the same. Is it correct?

Thanks for your help and any suggestions!

Best regards,

Yu

–
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 https://groups.google.com/group/gamsworld.
For more options, visit https://groups.google.com/d/optout.

2

Yu,

I think GAMS/Cplex never accepted the =c= syntax for SOCP problems. You always provided the SOCP via quadratic constraints. It was only GAMS/MOSEK and GAMS/CONOPT that understood the =c= notation. The =c= imposed many restrictions e.g. the check you have seen in your experiments. Even GAMS/MOSEK works not with providing the SOCP via a quadratic program. When you experiment with SOCP and Cplex, I recommend using the GAMS/CplexD link because is handles quadratic constraints in a better way (option qcp=cplexd).

Hope this helps,
Michael

On Saturday, January 16, 2016 at 12:01:48 PM UTC-5, yyang6@ualberta.ca wrote:

Hi all,

Sorry for the disturbance.

I am using the GAMS to solve the second order cone programming (SOCP). I model it by using =c= to specify the conic constraints and then I can use CPLEX to solve the resulting “LP” problem. I have several conic constraints and they share the same decision variables: x. The model seems correct but the GAMS shows that “Variable enters more than one =C= equation”. Does anyone know what’s going on here? Can I still use =c= equations (LP model) and solve the problem?

If I use the common NLP model and solve it by using CONOPT. Is there any difference in terms of the solving time or result compared with the “LP” model solved by CPLEX. This is convex so I think the result will be the same. Is it correct?

Thanks for your help and any suggestions!

Best regards,

Yu

–
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 https://groups.google.com/group/gamsworld.
For more options, visit https://groups.google.com/d/optout.

3

Many thanks!!

I am sorry that I reply this after nearly one month.

I simply use Baron to solve the QCP, but it seems not very efficient. I also tried CONOPT, but it told me that my model is locally infeasible! Then I followed your suggestion, using CPLEXD, and it works very well!!!

Best regards,

Yu

On Monday, January 18, 2016 at 10:10:53 AM UTC-5, Michael Bussieck wrote:

Yu,

I think GAMS/Cplex never accepted the =c= syntax for SOCP problems. You always provided the SOCP via quadratic constraints. It was only GAMS/MOSEK and GAMS/CONOPT that understood the =c= notation. The =c= imposed many restrictions e.g. the check you have seen in your experiments. Even GAMS/MOSEK works not with providing the SOCP via a quadratic program. When you experiment with SOCP and Cplex, I recommend using the GAMS/CplexD link because is handles quadratic constraints in a better way (option qcp=cplexd).

Hope this helps,
Michael

On Saturday, January 16, 2016 at 12:01:48 PM UTC-5, yya…@ualberta.ca wrote:

Hi all,

Sorry for the disturbance.

I am using the GAMS to solve the second order cone programming (SOCP). I model it by using =c= to specify the conic constraints and then I can use CPLEX to solve the resulting “LP” problem. I have several conic constraints and they share the same decision variables: x. The model seems correct but the GAMS shows that “Variable enters more than one =C= equation”. Does anyone know what’s going on here? Can I still use =c= equations (LP model) and solve the problem?

If I use the common NLP model and solve it by using CONOPT. Is there any difference in terms of the solving time or result compared with the “LP” model solved by CPLEX. This is convex so I think the result will be the same. Is it correct?

Thanks for your help and any suggestions!

Best regards,

Yu

–
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 https://groups.google.com/group/gamsworld.
For more options, visit https://groups.google.com/d/optout.