[gecode-users] Branching on domain size and picking constraints

Christian Schulte cschulte at kth.se
Tue Jun 22 15:58:02 CEST 2010


Hi,

No, that is not possible. But furthermore, it is not really clear to me why
you would pick a propagator? If the propagator needs to be run it will be
run. Otherwise, it will not.

Do you have some evidence that the default order of propagation is bad (or
even very bad) for an example you have? Or is it something you expect to be
bad?

If you really, really mean to exercise a very fine level of control over the
order of propagation you could use advisors for monitoring changes to
variables (see Chapter 16 in MPG) and in particular Section 16.4 on how to
force propagator re-scheduling.

Best
Christian

--
Christian Schulte, www.ict.kth.se/~cschulte/


-----Original Message-----
From: users-bounces at gecode.org [mailto:users-bounces at gecode.org] On Behalf
Of Peter Vanhee
Sent: Tuesday, June 22, 2010 12:06 PM
To: users at gecode.org
Subject: [gecode-users] Branching on domain size and picking constraints

Hi all,

I am wondering if a branching strategy can combine both

- variable picking according to their domain size (smallest first), can be
done using  INT_VAR_SIZE_MIN
- constraint picking, e.g. to chose the binary constraint attached to that
variable that connects with a variable
with a small domain as well

As an example, I have variables x, y & z, and domain sizes like x=10 y=10
z=1000. Binary constraints
are like x-y and x-z. My branching strategy picks a random variable of x and
then ideally would first propagate
this change to the domain of y. However, I saw it propagates to z first. How
can this be influenced, i.e.
how can the constraint x-y be forced to run before x-z, on the basis that y
has a smaller domain than z?

Thanks in advance!

Peter
_______________________________________________
Gecode users mailing list
users at gecode.org
https://www.gecode.org/mailman/listinfo/gecode-users




More information about the users mailing list