[gecode-users] BAB, constrain() and redundant cost constraints

[Peter Tiedemann]

On 14/04/07, Guido Tack <tack at ps.uni-sb.de> wrote:
> You usually work around this problem by adding another variable (say
> "y") to your model, adding the constraint "ax < y" to your model, and
> then just adding constraints "y < 10", "y<9" and so on during branch
> and bound. These constraints are not represented as propagators, but
> just immediately shrink the variable's domain.

I suspected i was missing something :) I hadnt noticed that that the
examples were posting constraints on a single variable. Thank you for
clarifying this for me.

> I think this model fits for most objective functions, but I would be
> very interested in an application where you can't avoid posting more
> and more real propagators during branch and bound.

Cant help you there, the objective i am considering is a simple
weighted sum of variable values, so the above trick fits nicely.

Regards,
Peter Tiedemann