Why MILP problem can not guarantee global optimal solution ?

GAMS
1

Hello,

I solved a MILP problem with CPLEX, and get a solution with final gap 0%. However, I use this solution as initial solution, resolve the problem, I can get a better solution. Does it mean the solution obtained earlier is just a locally optimal solution? I am quite confused why MILP problem can not guarantee global optimal solution.

Thanks a lot.

Best Regards,
Danya

–
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/d/optout.

2

if the gap is 0% it means the solution is optimal, check your better solution in all constraints and ensure that it satisfies all. I think your Heuristic algorithm doesn’t give the correct answer (if you used the cplex answer as an initial solution of a heuristic).

On Thursday, May 21, 2015 at 12:01:35 PM UTC+3, danya...@126.com wrote:

Hello,

I solved a MILP problem with CPLEX, and get a solution with final gap 0%. However, I use this solution as initial solution, resolve the problem, I can get a better solution. Does it mean the solution obtained earlier is just a locally optimal solution? I am quite confused why MILP problem can not guarantee global optimal solution.

Thanks a lot.

Best Regards,
Danya

–
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/d/optout.

3

By the way, use Baron solver also…

On Thursday, May 21, 2015 at 12:01:35 PM UTC+3, danya...@126.com wrote:

Hello,

I solved a MILP problem with CPLEX, and get a solution with final gap 0%. However, I use this solution as initial solution, resolve the problem, I can get a better solution. Does it mean the solution obtained earlier is just a locally optimal solution? I am quite confused why MILP problem can not guarantee global optimal solution.

Thanks a lot.

Best Regards,
Danya

–
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/d/optout.

4

Thanks Tino, the better solution satisfies all. My model is quite large with binary variable of 14,976, so it is difficult to get a solution using baron. In the case I descirbed before , I used the parallel mode of CPLEX by setting parallelmode 1 , threads 4. I am wondering is it possible that parallel affects the solution.

在 2015年5月21日星期四 UTC+8下午8:24:15，Tino写道：

if the gap is 0% it means the solution is optimal, check your better solution in all constraints and ensure that it satisfies all. I think your Heuristic algorithm doesn’t give the correct answer (if you used the cplex answer as an initial solution of a heuristic).

On Thursday, May 21, 2015 at 12:01:35 PM UTC+3, danya...@126.com wrote:

Hello,

I solved a MILP problem with CPLEX, and get a solution with final gap 0%. However, I use this solution as initial solution, resolve the problem, I can get a better solution. Does it mean the solution obtained earlier is just a locally optimal solution? I am quite confused why MILP problem can not guarantee global optimal solution.

Thanks a lot.

Best Regards,
Danya

–
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/d/optout.