[Scip] Solving a subproblem within a B&B node
Ambros Gleixner
gleixner at zib.de
Sat Apr 4 18:21:59 CEST 2015
Dear Andrei,
You are correct that changes in the sub-SCIP do not affect anything but
the sub-SCIP. Happy coding,
Ambros
Am 04.04.2015 um 03:41 schrieb Andrei Braga:
> Hello again,
>
> Reading the source code of SCIPcopy() method, I understood that it
> creates a
> new and independent SCIP instance, and copies into this instance information
> from the original one. Then, I concluded that my concern about changes in a
> sub-scip affecting the original SCIP instance makes no sense.
>
> I will keep going with my implementation, and report on further issues.
>
> Best regards,
>
>
> On Fri, Apr 3, 2015 at 11:41 AM, Andrei Braga <andreisampaio at gmail.com
> <mailto:andreisampaio at gmail.com>> wrote:
>
> Dear all,
>
> I am new to SCIP and I am trying to code a branch-and-price
> method. I'd like
> to ask you about how to implement the steps I describe below.
>
> I've read about "diving" and "probing", but I think these aren't
> approaches
> suited for my purposes. I guess I should use a "sub-SCIP" (which I
> suppose is
> obtained through a call to the scipCopy() method), but I don't
> understand the
> details of the process (for example, changes made in the sub-SCIP
> affect the
> original SCIP instance?).
>
> Let X be a subset of variables. When, after LP solving in a B&B
> node, all
> variables in X have integer value, I want to:
> 1) Create an copy of SCIP;
> 2) In this copy, delete some variables and constraints;
> 3) In this copy, replace the branching rules and the pricers by
> other ones;
> 4) In this copy, solve the "smaller" problem to optimality;
> 5) In the original SCIP instance, prune the B&B-node for optimality.
> I need that the deletions made in the SCIP copy do not affect the
> original
> SCIP instance.
>
> Thank you very much,
>
> --
> Andrei Braga
>
>
>
>
> --
> Andrei Braga
>
>
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> http://listserv.zib.de/mailman/listinfo/scip
>
--
______________________________________________________________
Ambros M. Gleixner
Zuse Institute Berlin - TU Berlin - Berlin Mathematical School
http://www.zib.de/gleixner
More information about the Scip
mailing list