Fw: user-defined (non-trivial) distribution strategy

[Thorsten Brunklaus]

----- Original Message ----- 
From: "Michael Richter" <xiaomir at yahoo.com>
To: "mozart_user_list" <users-request at mozart-oz.org>
Sent: Tuesday, September 03, 2002 5:56 PM
Subject: user-defined (non-trivial) distribution strategy


> hi,
> 
> i'm currently trying to reduce the size of a search
> tree. the distribution is made on a vector of
> variables with the domain 0#1, i have a cost function
> and i am only interested in the best solution. the
> optimal order of distribution is naive (due to
> propagation issues).
> 
> but i would like to speed up search in following the
> trail which optimises the cost function for current
> step like e.g. in the warehouse example (chapter 10 of
> FD constraint programming tutorial).
> unfortunately, the relation between the elements which
> i distribute to the cost function is only indirect,
> i.e. it needs some propagation unil relevant variables
> are changed. therefore it's not possible to use the
> straight approach (if this var bigger than another
> var...)
> 
> so my idea was to create both spaces (for 0 and for 1)
> and then decide to consider the one with better
> results first.
> 
> but, for now, i don't know 
> 1. how to do this with the predefined engines and
> 2. if i have to implement my own engine: how can i
> extract the information about the cost function from a
> space? (to use it for my decision?)
> 
> thanks in advance.
> 
> 
> with best regards,
> 
> michael richter
> 
> PS: some articles in the archive skim this question -
> but if there's an answer in there, i did not get it.
> 
> __________________________________________________
> Do You Yahoo!?
> Yahoo! Finance - Get real-time stock quotes
> http://finance.yahoo.com
> 

-
Please send submissions to users at mozart-oz.org
and administriva mail to users-request at mozart-oz.org.
The Mozart Oz web site is at http://www.mozart-oz.org/.