[SCIP] Quadratic Constraints

[Marc Pfetsch]

Hi Victor,

reoptimization can only be used if "the space of solutions gets
restricted and/or the objective function changes", see
https://www.scipopt.org/doc-7.0.0/html/REOPT.php

Note that "the objective function changes" only refers to linear
objectives, so I think that your case in which you are modeling a
quadratic objective function is not covered, but I agree that this would
be an interesting case.

So to achieve what you want, you cannot use reoptimization but need to
use the ordinary solving procedures. I should then be possible to change
the constraint.

Hope this helps.

Best

Marc


On 15/06/2020 22:19, Victor Miller wrote:
> I have a MIP that I want to solve many times, but each time with a
> different quadratic constraint of the form minimize sum_i  ||x||^2 -
> a^T x, where the vector a may change each time.  I understand that
> SCIP doesn't have quadratic constraints, but that I can specify that
> by introducing a continuous variable z, adding the constraint z >=
> ||x||^2 - a^T x, and minimizing z.
> 
> I'm using pyscipopt.  I do
> 
>         model.freeReoptSolve()
> 
>         model.chgReoptObjective(obj_var, "minimize")
> 
> However, since the constraint changes each time I want to delete the
> old constraint (if any),
> by doing
> 
> model.delCons(old_cons).
> 
> However, when I do that I get the error
> 
> Exception: SCIP: method cannot be called at this time in solution process!
> 
> How can I do what I want?  Is there a way of modifying the constraint
> instead of deleting it?
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> https://listserv.zib.de/mailman/listinfo/scip
>