[Scip] Best way to get the linear relaxation outside SCIPsolve(SCIP *) call?

[nikolaj@crt.umontreal.ca]

On Tue, May 26, 2009 9:05 am, Tobias Achterberg wrote:
> Hi Nikolaj,
>
>
>> Is there a possibility to access the optimal linear relaxation value of
>> an IP before calling SCIPsolve(SCIP *)? By calling the initial lpi (I'm
>> working with CLP) and calling CLP to solve it for example?
>
> Unfortunately, there is no convenient way to do this. The LP relaxation
> of the CIP problem only exists during the solving stage, i.e., inside
> SCIPsolve(). What you could do is to
> change all variables into continuous variables, call SCIPsolve(), query
> the solution, call SCIPfreeTransform(), and then change them back to
> integers.
>
> Note that the LP relaxation value obtained in this way will usually be
> different from the root LP value of SCIP, because SCIP is exploiting
> integrality in presolve to tighten the formulation.
>
>
>> Same question but after SCIP has found an (integer valued) optimal
>> solution and exited the SCIPsolve(SCIP *) method.
>
> I don't understand this second question. Do you want to solve the MIP
> with SCIPsolve() and afterwards solve the LP relaxation of the MIP? In
> this case, you can do the same as above: SCIPsolve(), query solution,
> SCIPfreeTransform(), just change all variables into
> continuous variables and call SCIPsolve() again.
>
>
> Tobias
>
>

Hi Tobias,

Thank you very much for your answer. I'll do what you suggested.

About my second question : I just wanted to save myself additionnal
calculations but as you mentionned it, the LP integer relaxation at the
root of the search tree is not the same as the LP integer relaxation of my
model.

Have a nice day,

Nikolaj