I have some very large MIPs to solve. I can relax integrality conditions on some variables but realized that doing so makes the root node solve time explode. I guess Gurobi applied cuts and strengthens bounds/coefficients based on the integrality conditions, making the LP relaxation easier to solve at first (with integrality conditions).
Can I solve the root node with integer variables, and after that change the type of some variables, continuing the branching phase with fewer integers, but still benefiting from new cuts/bounds?
Please sign in to leave a comment.