I have a class of MILP problems for which I know the solution a priori. As such, I know that gurobi finds the optimal solution relatively early, but the problem is large and thus long time is spent proving optimality.
I was thinking of writing a callback that checks if the solution has changed for N nodes and/or T seconds (based on some rule of thumb I have). For large T or N after which the best found solution is the same, I would assume that this is optimal or near optimal. At this point, I can terminate the solving process. But instead of terminating, are there parameters that I can change or strategies that I can exploit so that gurobi stops focusing on finding a better solution and proves optimality faster?
Please sign in to leave a comment.