Topic
  • 2 replies
  • Latest Post - ‏2013-01-07T04:21:37Z by Uonly
Uonly
Uonly
16 Posts

Pinned topic setting upper bound for MIP

‏2013-01-05T07:40:36Z |
For a minimization problem, I want to set an upper bound. How to do that?

How did Cplex handle this? Add obj <= ub? as a constraint?

If the the upper bound was set to tight, the feasible problem can be wrongly claimed to be infeasible.

What will the message CPLEX outputs? Infeasible problems?

I understand users can feed any feasible solutions into Cplex from any heuristics but here I am considering arbitrary global upper bound.

Thanks
Updated on 2013-01-07T04:21:37Z at 2013-01-07T04:21:37Z by Uonly
  • SystemAdmin
    SystemAdmin
    7929 Posts

    Re: setting upper bound for MIP

    ‏2013-01-06T17:53:27Z  
    > Uonly wrote:
    > For a minimization problem, I want to set an upper bound. How to do that?
    > How did Cplex handle this? Add obj <= ub? as a constraint?
    >
    This bound can be set using parameter CPX_PARAM_CUTUP.
    CPLEX will cut off all nodes for which the node's objective function value exceeds this cutoff.

    > If the the upper bound was set to tight, the feasible problem can be wrongly claimed to be infeasible.
    >
    Correct.

    > What will the message CPLEX outputs? Infeasible problems?
    >
    Yes, CPLEX will report the problem infeasible (since no feasible solution was found).
  • Uonly
    Uonly
    16 Posts

    Re: setting upper bound for MIP

    ‏2013-01-07T04:21:37Z  
    > Uonly wrote:
    > For a minimization problem, I want to set an upper bound. How to do that?
    > How did Cplex handle this? Add obj <= ub? as a constraint?
    >
    This bound can be set using parameter CPX_PARAM_CUTUP.
    CPLEX will cut off all nodes for which the node's objective function value exceeds this cutoff.

    > If the the upper bound was set to tight, the feasible problem can be wrongly claimed to be infeasible.
    >
    Correct.

    > What will the message CPLEX outputs? Infeasible problems?
    >
    Yes, CPLEX will report the problem infeasible (since no feasible solution was found).
    thanks Daniels