Originally posted by: jacobmays

Hi,

I have implemented an L-shaped method in AMPL, using CPLEX 12.6.2.0. The master problem is a MIP, while the subproblems are all LPs.

In the later iterations, I consistently find that the output of the subproblems is:

CPLEX 12.6.2.0: optimal solution; objective #####
0 simplex iterations (0 in phase I),

where the ##### is the objective function value of the given subproblem and changes every iteration.

My question is: given that the solution to the master affects only the constraints of the subproblem, how does CPLEX produce a new solution (as evidenced by the changing objective function value) with 0 iterations?

Thanks,

Jacob

Originally posted by: BoJensen

If the constraints of the problem has changed and there's a basic solution stored in CPLEX, then the stored basic solution will be recalculated and may still be optimal, but the numerical values of the solution and it's objective value has changed without any iteration has been taken. 

You can check if this happens in your case by writing out the solution and problems to file and load it into the interactive command prompt.