MIP in an iterative method

Dear members,

I have a MIP mathematical model, I want to solve it for many times. For example, the model is

Min CX

s.t.

AX<=B

After each iteration, the parameter B is changed. I do not want the new model to solve again. I know the solution can be found from the previous branch and bound tree.How can I use the previous branch and bound tree and nodes for the new problem?


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.

Hi,

General purpose MIP solvers (as available in the GAMS) cannot reuse the B&C tree from previous runs to improve performance. This is not a lack of functionality of these solvers but in general a very hard to do. While there is some theory (see e.g. http://www.lehigh.edu/~tkr2/research/talks/Duality05.pdf) this has not penetrated the commercial MIP solver world. The only thing you can do it to restart your B&C with an (almost) feasible solution (solver option “mipstart”). Some solvers try to repair an infeasible MIP start, but that helps only with the primal bound, the solver needs to branch-and-cut from scratch to solve the problem to (near) optimality.

Hope this helps,
Michael Bussieck - GAMSWorld Coordinator

On Thursday, August 14, 2014 6:14:14 AM UTC-4, hossein karimi wrote:

Dear members,

I have a MIP mathematical model, I want to solve it for many times. For example, the model is

Min CX

s.t.

AX<=B

After each iteration, the parameter B is changed. I do not want the new model to solve again. I know the solution can be found from the previous branch and bound tree.How can I use the previous branch and bound tree and nodes for the new problem?


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.

Thanks Michael,
I think the “mipstart” is not useful for my problem.


On Thu, Aug 14, 2014 at 7:53 PM, Michael Bussieck wrote:

Hi,

General purpose MIP solvers (as available in the GAMS) cannot reuse the B&C tree from previous runs to improve performance. This is not a lack of functionality of these solvers but in general a very hard to do. While there is some theory (see e.g. http://www.lehigh.edu/~tkr2/research/talks/Duality05.pdf) this has not penetrated the commercial MIP solver world. The only thing you can do it to restart your B&C with an (almost) feasible solution (solver option “mipstart”). Some solvers try to repair an infeasible MIP start, but that helps only with the primal bound, the solver needs to branch-and-cut from scratch to solve the problem to (near) optimality.

Hope this helps,
Michael Bussieck - GAMSWorld Coordinator


On Thursday, August 14, 2014 6:14:14 AM UTC-4, hossein karimi wrote:

Dear members,

I have a MIP mathematical model, I want to solve it for many times. For example, the model is

Min CX

s.t.

AX<=B

After each iteration, the parameter B is changed. I do not want the new model to solve again. I know the solution can be found from the previous branch and bound tree.How can I use the previous branch and bound tree and nodes for the new problem?


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.

\

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.