Hello everybody,
I'm new to optimization and GAMS and I've got following problem:
When using Gurobi for a large MIPProblem and uploading the code to https://neosserver.org
i receive different Outputs for the value "best possible" for different optcr settings.
I know why I receive different results for "Final Solve" and I'm not able to set optcr to 0.0 because the problem is to large to solve it in 8 hours, but i cannot understand why the best possible solution differs as the option optcr should not affect the real optimum?
Is the reason for the differeing best possible values the progress Gurobi made in branching and bounding, when finally solving the problem?
Thanks!
Output Gurobi, Best possible

 User
 Posts: 85
 Joined: 10 months ago
Re: Output Gurobi, Best possible
You didnot get the real optimum when optcr is not 0.
optcr is relative termination tolerance for use in solving MIP problems.
even optcr is 0, you still get different optimal results because for the same optimization problem there could be multiple optimal solutions which will result in the same objective value.
I suggest you try to understand how an optimization problem is solved in mathematics.
optcr is relative termination tolerance for use in solving MIP problems.
even optcr is 0, you still get different optimal results because for the same optimization problem there could be multiple optimal solutions which will result in the same objective value.
I suggest you try to understand how an optimization problem is solved in mathematics.
Re: Output Gurobi, Best possible
Thanks for your answer.
But as I said, Im not confused by receiving a solution that is not optimal. I'm confused by the fact that the best possible solution in the output changes.
best possible  final solution = gap (=optca)
But as I said, Im not confused by receiving a solution that is not optimal. I'm confused by the fact that the best possible solution in the output changes.
best possible  final solution = gap (=optca)