[gecode-users] Skipping branching alternatives?

[Christian Schulte]

I do not know whether BAB will be good enough as I do not know what Martin
exactly needs. However I would also supsect that it will be BAB + X, where X
might be to get more information from search so far.

Christian

--
Christian Schulte, http://www.imit.kth.se/~schulte/ 

-----Original Message-----
From: Luis Quesada [mailto:luque at info.ucl.ac.be] 
Sent: Tuesday, July 11, 2006 10:04 AM
To: schulte at imit.kth.se
Cc: 'Martin Mann'; 'gecode user list'
Subject: Re: [gecode-users] Skipping branching alternatives?


Christian Schulte wrote:

>I just reply to this email, even though I have read the discussion. I 
>very much second Guido and Mikael and I have another bit to add as to 
>why your idea of tinkering with branchings does not work.
>
>The entire setup of search in Gecode relies on recomputation: here, a 
>commit operation using a branching description _must_ always tell the 
>very same constraint when invoked with the same parameters. So by 
>definition, it is not allowed to use state in the way you state it.
>
>  
>
I confess that I still don't see why you need to implement another 
search engine for dealing with Martin's problem. It seems to me that BAB 
would do the job....Cannot Martin do what he wants to do by using the 
constrain method? To me, that's the best place to address his global 
property.

Luis