Re: CSP Platform?
- From: "Paul A. Rubin" <rubin@xxxxxxx>
- Date: Thu, 25 Oct 2007 10:45:24 -0400
A.L. wrote:
Actually, CP is not MP and having experience with CP I have, I
simply don't believe that it is possible to treat CP solver as
"black box" to the same extent as LP solver. Predefined constraints
predefined constraints, but strategy of labeling and augmenting
standard propagating by business based propagator, symmetry removal
and such seems to be critical.
I wondered about that during the webinar, but chalked it up to my lack of background in CP. Maybe this is an instance of the 80-20 rule: 80% of potential CP problems use 20% of the possible constraints (things like all-different), and the other 20% of potential CP problems are a royal PITA and need custom coding. Even in MP, I've become dependent on callbacks and customized algorithms, but most LP models can be solved easily with any good solver and default parameter settings.
You can write a sudoku solver using just common constraints and whatever the standard propagation methods are, right? What more can you ask? :-)
/Paul
.
- References:
- CSP Platform?
- From: Malcolm Ryan
- Re: CSP Platform?
- From: A . L .
- Re: CSP Platform?
- From: Paul A. Rubin
- Re: CSP Platform?
- From: A . L .
- Re: CSP Platform?
- From: Paul A. Rubin
- Re: CSP Platform?
- From: A . L .
- Re: CSP Platform?
- From: Paul A. Rubin
- Re: CSP Platform?
- From: A . L .
- CSP Platform?
- Prev by Date: Re: CSP Platform?
- Next by Date: Re: CSP Platform?
- Previous by thread: Re: CSP Platform?
- Next by thread: Re: CSP Platform?
- Index(es):
Relevant Pages
|
|