[gecode-users] Intelligent backtracking methods
Tias Guns
tias.guns at cs.kuleuven.be
Tue Jun 10 14:27:45 CEST 2008
Just out of interesest, I wonder how this relates to the constraint store ?
More specifically, if a constraint does GECODE_REWRITE(...) I can imagine
the space could be cloned at that point ? Anything further in the search
space will be recomputed from that clone (with rewritten constraint) on,
and backtracking to something before will be recomputed from an earlier
clone (with original constraint) ?
Or is this independant, and will a recomputation (from a clone before the
rewrite) just rewrite the constraint over and over again (for different
recomputations) ?
Given the modular way Gecode is written, this latter seems rather
plausible. It's just a bit surprising that would be so efficient...
Greetings,
Tias
On Tue, 10 Jun 2008 10:18:14 +0200, Christian Schulte <cschulte at kth.se>
wrote:
> Not that much dedicated to Gecode in particular. But the interface is in
> large aspects very close to the Mozart interface for search which is well
> documented:
> http://web.it.kth.se/~cschulte/paper.html?id=Schulte:LNAI:2002
>
> The only difference is how recomputation in detail works, but that
> becomes
> rather obvious when checking how recomputation is implemented for the
> search
> engines in Gecode (check the file reco-stack).
>
> Cheers
> 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 Debdeep Banerjee
> Sent: Tuesday, June 10, 2008 9:34 AM
> To: cschulte at kth.se; users at gecode.org
> Subject: Re: [gecode-users] Intelligent backtracking methods
>
> Hi,
>
>> However, you can implement your own search methods as Gecode features an
>> enormously powerful interface for programming search. In particular, as
>> search is based on recomputation and copying, any move in the tree is
> easy.
>>
> Is there any paper or documentation for that recomputation stuff ?
>
> with regards
> Debdeep
>
More information about the gecode-users
mailing list