I am trying to solve a problem with Branch-and-Bound(Off parameters 'Heuristics', 'Presolve', 'Symmetry', 'Cuts') for the Binary Integer Programming model studying now
I'm currently interested in Cutoff and have two questions regarding it.
1. For my model(aim to minimize), I tested re-solving model(10 instances) by setting the cutoff value as the optimal value for the problem already solved with Branch-and-Bound(B&B). In this case, the time it took to solve the problem actually increased. Even if the solution was found at similar time as before, time to completion was slower.
I know if setting an upper bound, it should cut off values that are worse than that value from the branch-and-bound tree, reduce nodes that need to be explored. Could you advise me why this is happening?
Branch and Bound (computation time)
8.59 6.39 8.32 5.77 8.99 32.38 7.63 13.82 26.49 23.09
Adding Cutoff(Cutoff value = Optimal Value) parameter (computation time)
32.62, 11.11, 11.80, 5.47, 10.27, 28.06, 10.85, 30.75, 30.27, 17.92
2. Could you advise me when I get a lower bound on the LP relaxation objective value in a Branch-and-Bound tree using heuristics and if objective value of the LP relaxation is less than this, is there a way to cut off these nodes in the tree?
Please sign in to leave a comment.