Originally posted by: Itherain

Hi all,

  I need to solve a lots subproblems in my algorithm. I set the best upper bound to the subproblem and I use the parameter CPX_PARAM_CUTUP to cut some useless nodes. However, I have no ides about the difference between give an upper bound (objective value) and give its corresponding solution.

Thanks for any reply.

With CPX_PARAM_CUTUP you basically promise that the optimal solution is better than that value. CPLEX does not verify that information, it just trusts you.

Instead of using CPX_PARAM_CUTUP you could also provide your solution as a MIP start. That incurs a little more overhead but has some advantages:

• CPLEX actually has a feasible solution with the specified value, so you can be sure that you don't cut anything wrong
• CPLEX also checks your solution for feasibility, so you can recognize any errors
• CPLEX can run improvement heuristics on that solution, this may potentially speed up the solution process