Christian Schulte wrote:
> How about the following generalization: one can pass to a brancher a filter
> function (that is, a member function of the space). If a filter function is
> passed it is used for selecting variables for branching: only if the filter
> function returns true for a variable, the variable is considered for normal
> selection. The function can be passed through the normal options for
> branching (VarBranch) that is also used for tie breaking and randomized
> selection.
>
> Would that make sense? It is less general than what Guido proposed but
> easily available to the average modeler. I thought about it a little and I
> know how to efficiently implement it.
The filter function could get the index of the variable in addition to the
variable itself. That would make it just as general as my approach, I think
(you could use a different interval for each variable without having to go
through the whole array to find out which variable you are dealing with).
Cheers,
Guido
--
Guido Tack, http://people.cs.kuleuven.be/~guido.tack/
_______________________________________________
Gecode users mailing list
[email protected]
https://www.gecode.org/mailman/listinfo/gecode-users
