Re: [eclipse-clp-users] Guess Free Solution

From: Helmut Simonis <H.Simonis_at_4c.ucc.ie>
Date: Tue, 9 Feb 2010 18:02:14 -0000
Alex,

 

this is just what propagation does. If you state all constraints, before starting a search , you can check the domains of the variables to see which values have been removed by propagation. 

 

Sometimes, your constraint model is not strong enough to find all inconsistent values, in which case you can make stronger propagation by using a different model or by implementing better reasoning inside some constraint.

 

You may want to look at my paper on Sudoku or the one on Kakuro  (http://4c.ucc.ie/~hsimonis) to see how you can strengthen a model to deduce more and more information without search. Unfortunately, this works very well on puzzles, but not so well on other problem types.

 

Helmut 

 

From: Alex Rozenshteyn [mailto:rpglover64_at_gmail.com] 
Sent: 09 February 2010 17:48
To: Wit Jakuczun
Cc: eclipse-clp-users_at_lists.sourceforge.net
Subject: Re: [eclipse-clp-users] Guess Free Solution

 

Using sudoku as an example, only place a number if through examining the board it can determine that the number must fill that particular space, either because no other number can or because no other square can hold that number or for some other reason that can be explained as a deterministic step.

The idea behind this is that I'm trying to determine if a puzzle is solvable in a way such that for every step in the solution, there exists a logical next step to take.

On Tue, Feb 9, 2010 at 12:22 AM, Wit Jakuczun <wit.jakuczun_at_gmail.com> wrote:

2010/2/8 Alex Rozenshteyn <rpglover64_at_gmail.com>:

> Is there a way to tell eclipse to "never guess", i.e. only fill a constraint
> if there is a "logical" reason to?
>

 I do not understand what you mean by "never guess" and "fill a
constraint"? Could you elaborate on this?


Best regards
--
[ Wit Jakuczun  http://wlogsolutions.com ]




-- 
         Alex R
Received on Tue Feb 09 2010 - 18:03:02 CET

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