Bound update rule

Can Li

• February 21, 2020 01:00

Hi folks,

 

Suppose I have a minimization problem (an MILP) to solve. I wonder if there is a way to change how Gurobi updates its upper bound. 

 

To be more specific, in the original implementation, gurobi will try to update the upper bound once it finds a feasible solution at a given node. But now I want to prevent gurobi from updating the upper bound based on feasible solutions of the MILP. Instead, I want to define a lazycallback such that whenever gurobi finds a feasible solution, in the callback I will evaluate the solution and come up with a "real value" of this solution, which is different than the value of this solution in the MILP. I want to use that "real value" in the callback to update the upper bound of the MILP. I wonder if this is doable in Gurobi.

 

Thanks in advance!

 

  

0

• 

  Yuriy Zinchenko

  Gurobi Staff

  • February 24, 2020 21:40

  Can you add your own cut in the callback, using the newly-determined value for the objective cut-off?

  0
• 

  Can Li

  • February 24, 2020 21:49

  

  

  

  Hi Yuriy,

   

  Thanks for your reply. The thing is that the "real value" is higher than the upper bound found by the solver. I don't know how to prevent the solver from updating the upper bound with the incumbent solution of the MILP. I'm not very sure what you mean by "objective cut-off". The "real value" can only provide an upper bound for the minimization problem.

  0

Please sign in to leave a comment.