Would it be possible for you to help me by expanding the information about the meaning of the following states?
102 optimal (non-)integer solution 103 optimal (non-)integer solution within mipgap or absmipgap 110 optimal with unscaled infeasibilities 111 integer optimal with unscaled infeasibilities
What do these statuses imply? What does unscaled infeasibilities mean?
These states indicate that the solver has reported an optimal solution, but that the solution might not be as expected for some reason.
102 optimal (non-)integer solution
103 optimal (non-)integer solution within mipgap or absmipgap
Some of the variables that are defined as “integer” or “binary” may have slightly non-integer values, like 0.99999 or 5.00001, in the optimal solution. (If the gap between the lower and upper bounds at the optimal solution is not zero, then also “within mipgap or absmipgap” appears.) Some solvers have options to force all integer variables to exactly integer values, though that can require extra work.
110 optimal with unscaled infeasibilities
111 integer optimal with unscaled infeasibilities
Before solving, the rows and columns of the constraint coefficient matrix may be scaled – divided by some positive numbers – because scaling often gives better solver performance. The optimal solution to this scaled problem is allowed to violate some of the bounds and constraints, if the violation is less than a small feasibility tolerance (typically 0.0001). At the end of the solver run, the scaling is reversed, to provide a solution to the original, unscaled problem. It can happen that the solution to the unscaled problem violates some bounds or constraints by more than the solver’s feasibility tolerance, in which case the “unscaled infeasibilities” state is reported. Some solvers have options to reduce scaling or tighten the feasibility tolerances, which might help with this situation.
I say “some solvers” in my comments because solvers vary in which options they provide. Lists of options are provided on our Solvers page.