[gecode-users] value-consistency definition

[Mikael Zayenz Lagerkvist]

On 5/3/07, Guido Tack <tack at ps.uni-sb.de> wrote:
> > I've searched for a definition of value-consistency, but I don't
> > really find it.
>
> Unlike bounds-consistency or domain-consistency, it's not standard
> terminology, I think.

The closest theoretical notion is probably forward checking.

However, it's good to remember that forward checking may be defined in
several ways for non-binary constraints [1]. The idea of only
subscribing to PC_INT_VAL events, but doing full domain consistent
propagation (which roughly corresponds to some of the suggestions in
the mentioned paper) is not a correct approach for Gecode, since this
would make the propagator non-monotonic. See the recent discussion
with Peter Tiedemann about non-monotonic propagators for some more
information [2].


Cheers,
Mikael


[1] "On forward checking for non-binary constraint satisfaction", C.
Bessiere, P. Meseguer, E.C. Freuder and J. Larrosa, Artificial
Intelligence, volume 141, pages 205-224
[2] http://www.gecode.org/gecode-users/2007-April/001481.html

-- 
Mikael Zayenz Lagerkvist, http://www.ict.kth.se/~zayenz/