Good morning and thanks for your Answer first,
I already read the Section B, my problem is that i still want to use the
variable-based branchers within my own brancher.
My question therefore is how can i add a local instance of a brancher in
my self-written brancher?
As i understood i cannot call branch(...) inside my own brancher.
Because this would register the brancher on the Space.
I need a way to instanciate the brancher and then call the 'child'
brancher in choice/commit/(status) function from my brancher.
Can you give me a hint howto create a brancher for this?
I tried to change the brancher for my space, but unfortunately the
brancher(id) function are pure private to space. Therefore it seems to
forbidden to do this on the space directly.
Regarding AFC, as i understood it does not change the layout of the
search tree (which on the other hand would be the case if i write a own
brancher). It only searches for another node to continue the search. I
Still want all solution for my 'problem'. Therefore AFC might (from my
understanding) not the way to go?
Best,
Matthias
On 13.04.2015 21:32, Christian Schulte wrote:
Hi Matthias,
Well there is a very simple answer which might be not very satisfying, though:
the variable-value based branchings are just some pre-defined abstractions that
come with Gecode. In principle you can implement whatever you like, the entire
Part B of MPG is devoted to this.
I do not know how much you have been exploring with the pre-defined variable
selection strategies. But maybe you want to try ideas such as AFC and using
restarts for search first. They sometimes can be remarkably effective.
Cheers
Christian
--
Christian Schulte, www.gecode.org/~schulte
Professor of Computer Science, KTH, cschu...@kth.se
Expert Researcher, SICS, cschu...@sics.se
-----Original Message-----
From: users-boun...@gecode.org [mailto:users-boun...@gecode.org] On Behalf Of
Matthias Goldhoorn
Sent: Monday, April 13, 2015 04:04 PM
To: users@gecode.org
Subject: [gecode-users] Dynamic Branching Support?
Hi to everyone on the ML,
i'm new to gecode and have already a kind of bug problem .
My Search will not Terminate within days. I have several SetVarArrays,
BoolArrays and so on on my Space.
The Solutions they are found are valid for me, but the search is simply to slow
(and sometimes to memory exhaustive).
I used already gist for introspection, my problem is that i first branch on a
BoolVarArrays which reduces the search space quite well until a spefici point.
At this point i would like to keep Post prune the BoolVarArray and Continue
branching on a SetVarArray. This wold faster reduce my search space than
continue the branching on the Bool one.
My question therefore is: is there a way to select dynamically the branching
set?. I found on the documentation that it is possible to select a variable
INSIDE a structure for branching. But that i need is a dynamic selection of the
Set/Array where the branching should be done.
Is this possible in gecode. Or (@dev's) is the extension possible?
Best,
Matthias
(P.S. Since i do open-source development i could provide my application if this
helps the understanding.)
--
Dipl.-Inf. Matthias Goldhoorn
Space and Underwater Robotic
Universität Bremen
FB 3 - Mathematik und Informatik
AG Robotik
Robert-Hooke-Straße 5
28359 Bremen, Germany
Tel.: +49 421 178 45-4193
Zentrale: +49 421 178 45-6550
Fax: +49 421 178 45-4150
E-Mail: matthias.goldho...@uni-bremen.de
Weitere Informationen: http://www.informatik.uni-bremen.de/robotik
_______________________________________________
Gecode users mailing list
users@gecode.org
https://www.gecode.org/mailman/listinfo/gecode-users
--
Dipl.-Inf. Matthias Goldhoorn
Space and Underwater Robotic
Universität Bremen
FB 3 - Mathematik und Informatik
AG Robotik
Robert-Hooke-Straße 5
28359 Bremen, Germany
Tel.: +49 421 178 45-4193
Zentrale: +49 421 178 45-6550
Fax: +49 421 178 45-4150
E-Mail: matthias.goldho...@uni-bremen.de
Weitere Informationen: http://www.informatik.uni-bremen.de/robotik
_______________________________________________
Gecode users mailing list
users@gecode.org
https://www.gecode.org/mailman/listinfo/gecode-users