[SCIP] Questions regarding B&P framework for max problem.

[Felipe Serrano]

Hi

Yes, SCIP copies the problem, changes it to a minimization problem and
solves the minimization version. Also be aware of the warning in the
documentation of SCIPcreateVarBasic:
https://scipopt.org/doc/html/group__PublicVariableMethods.php#ga4a613346a942e7f962a69ed3568ce72d

Best,
Felipe

On Wed, Dec 2, 2020 at 10:01 AM Shin, Jungeun <jungeun4 at illinois.edu> wrote:

> Dear SCIP users
>
> I am trying to solve maximization problem using B&P framework and found a
> few things I can’t understand from my knowledge.
>
> My problem is defined as follows,
>
> Max cx
> Ax=1     ——— constraint (1)
> Bx - Dy <=0 ——constraint (2)
> 0<= x <= 1
> y binary
>
> Pricing problem generates x columns and here are some questions that I
> have.
>
> 1) I thought that its dual problem should give upperbound and the dual
> multiplier of (2) is non-negative.  However, getDualsolLinear(2) returns
> negative values.
>
> 2) The cost vector c has all positive elements. When I try to get the lp
> relaxation value within the Pricer using getLPObjVal( -0.115) , it returns
> negative value, while the primalbound is positive value(0.106).
>
> If I understood the primal-dual relationship correctly, from the above two
> observations, I suspect if the SCIP framework automatically convert
> maximization problem to minimization problem by changing cost
> vector signs ??
>
> I might have made some mistakes while writing the model. I check that the
> problem is set to maximize. I can’t find other possibilities that could
> make this happen.
> I would be grateful for any comment.
>
> Thank you!
>
>
>
>
>
>
>
>
>
>
>
>
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> https://listserv.zib.de/mailman/listinfo/scip
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.zib.de/pipermail/scip/attachments/20201202/30214b75/attachment.html>