Originally posted by: SystemAdmin

[Bill said:]

Hey,

I'm using barrier algorithm to solve (a huge) LP problem without using crossover step (because it is time consuming). I need the dual multipliers to solve another problem, so my question is : is the vector I get when I use IloCplex::getDuals() function an approximation of the simplex dual multipliers ? (I'm not sur, but I think no. Please correct me if I'm mistaken).

If it is not the case, how can I have access to the dual multipliers (if it's possible).

Thanks for help.

Cheers,
Bill

I'm using Cplex 11 using C++ API;
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

[teresa said:]

I don't have an answer to your question, but, as you are using barrier, i wonder if you can give me a tip on how to overcame lack of memory problems (i have already tried Emphasys memory parameter)
Thank a lot, Teresa
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

[achterberg said:]

The dual multipliers that you get from barrier have the same meaning as for the simplex code. That is, the dual vector is a feasible solution to the dual of your LP. In other words, if you aggregate the rows of the problem using the weights of the dual vector, you will get a proof of optimality.

The difference to a simplex solution (for both, the primal and the dual solution vector) is just that the barrier solution is typically not a vertex of the primal and dual LP polyhedra. In other words, it does not correspond to a basis.

#CPLEXOptimizers
#DecisionOptimization