Skip to main content

Does the Gurobi MIP Solver use safe bounds?



1 comment

  • Jaromił Najman
    Gurobi Staff Gurobi Staff

    Hi Renan,

    Gurobi will always try to be numerically as accurate as possible even for numerically questionable models. However, for numerically difficult models, Gurobi cannot guarantee 100% correctness due to computational tolerances and machine accuracy. For your particular example, it is very likely that Gurobi will perform the correct computations but it is not guaranteed due to numerical instability of the model. You could try experimenting with the parameters: FeasibilityTol, IntFeasTol, OptimalityTol, NumericFocus, and IntegralityFocus to increase Gurobi's computational accuracy.

    In cases like this, it is always best to improve the numerical properties of your model to fight the source of the issue instead of only fighting the symptoms via parameter settings. You can find more information about tackling numerical issues in our Guidelines for Numerical Issues.

    Best regards, 


Please sign in to leave a comment.