Here's the definition of MIPGap used in Gurobi:
It says that:
gap = ( ObjBound - ObjVal ) / | ObjVal |
where ObjBound and ObjVal are the MIP objective bound and incumbent solution objective, respectively.
If the problem is of the minimization type, won't this give a negative gap? For example, if the best LB is 5 and the best UB (given by incumbent solution) is 10, it should say gap = ( 5 - 10 ) / |10| = - 0.5, or - 50%.
However, Gurobi does not report a negative gap when I solve minimization problems. What am I missing?
Thanks,
Austin