[Scip] Can I tell SCIP to refrain from fixing/aggregating some variables?

[Timo Berthold]

Hi Daniel.

Do you add constraints 2) and 3) as cuts (SCIPcreateRow(), SCIPaddCut) or
as linear constraints (SCIPcreateConsLinear(), SCIPaddCons()) ? For the
difference between Constraint handlers and separators, see
http://scip.zib.de/doc/html/FAQ.html#Q4.3

The second would be the correct way. The plugin adding them should be a
constraint handler itself. This constraint handler then needs to set locks
on the variables y (and in principle also x_i) to avoid dual fixing
mechanisms to be applied. Also, see http://scip.zib.de/doc/html/CONS.html

Cheers,
Timo

> Hi everyone,
>
> I have the following problem. In my model I have continuous variables x0
> andy
> , binary variables x1, ..., xk, and linear constraints of the following
> form:
>
> 1)         x1 +    x2 + ... +    xk = 1
> 2) x0 + a1*x1 + a2*x2 + ... + ak*xk = y
> 3) lb <= x0 <= ub
>
> There are other constraints, too, but they do not matter here.
> The objective is to minimize y.
> The idea is that I can write y as
>
> y = ai + x0,
>
> for some i so that I can branch on the variables xi during the algorithm.
> But I don't know the coefficients ai and the bounds lb and ub beforehand,
> and the constraints 2) and 3) are added later on, once I have determined
> the missing pieces.
> Now SCIP does not know that x0 will become important later on and fixes it
> to zero.
> Is there a way to exclude the variables xi from the presolving or to
> explicitly forbid fixing or aggregating them?
>
> Best regards,
>
>   Daniel
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> http://listserv.zib.de/mailman/listinfo/scip
>