Originally posted by: disjunctive

Hello,

I am using Java API and I am trying to solve a two-stage MIP. At each iteration, both two stages are added cuts. However, I am not using lazy constraints. I just add constraints (cuts) to the LPMatrix instances. I do not know this approach is correct. Because the solution is not correct. Do you think I need to use lazy constraints? 

Also, I have another question. Second stage problem is MIP too but I am solving the LP relaxed problem. When I am creating the decision variables, I predefine them as continuous variables. I have some suspicions. If someone clear the air, I would appreciate.

Thanks in advance

First question: If you are solving both problems to convergence in each iteration (rather than using callbacks), then you do not need lazy constraints. If you think that cuts generated during early iterations may become redundant later in the process, it might be faster to add them as lazy constraints, though, so that they don't clutter up the active matrix. If you are using callbacks, and the cuts can remove otherwise feasible solutions, then you need lazy constraint callbacks.

Second question: If all you care about is the LP relaxation, using continuous variables should be fine.