Re: [eclipse-clp-users] RES: RES: RES: minlist /2 in a Point list

From: Wit Jakuczun <wit.jakuczun_at_gmail.com>
Date: Fri, 25 Jun 2010 13:47:57 +0200
W dniu 2010-06-25 01:38, Igor Kondrasovas pisze:
> Hello Joachim,
>
>    
>> Only after the constraints have done their work ("propagated the
>>      
> consequences of the search decision"), and only of it didn't lead to
> failure, is a new>child search node created.  The whole point of
> constraint-based search is that you do NOT have to go all the way down to
> the leaves of a search tree!
>
> In my understanding I see the search tree of my (and many others) problem
> having each variable (X1, Y1, X2, Y2, etc) at a distinct level of the search
> tree, like shown in Page 124 of the Eclipse Tutorial. So a candidate
> solution of my problem would only be found when the search tree reaches the
> final node (a leaf). This way, what am I missing out?
>    
You are in general right by saying that "

So a candidate
solution of my problem would only be found when the search tree reaches the
final node (a leaf).

" but clp approach gives you a bit more because after you choose some 
value for eg. variable X1, then by propagation you can cut-off some part 
of the search tree (eg. X1=10 => X2 \= 10). So if you have good 
propagation then you will have to check only a part of the whole search 
tree.

Best regards

-- 
[ Wit Jakuczun http://pl.linkedin.com/in/jakuczunwit ]
[ WLOG Solutions http://wlogsolutions.com            ]
Received on Fri Jun 25 2010 - 11:48:08 CEST

This archive was generated by hypermail 2.2.0 : Thu Feb 02 2012 - 02:31:58 CET