Re: Assignmnet problem with rules
- From: Logan Shaw <lshaw-usenet@xxxxxxxxxxxxx>
- Date: Tue, 28 Feb 2006 03:22:11 GMT
Willem wrote:
Daniel wrote:
) Come up with initial assignments that don't break the independent rules, ) then check to see how many of the dependent rules are broken. Start ) swapping assignments, if fewer rules are broken, keep the swap, if more ) rules are broken swap back. The longer the program runs, the fewer rules ) that will be broken, but you can get a solution at any time.
This way, you'll only reach a local minimum. You could add a variable that
indicates the probability of continuing with a 'worse' swap, and decrease
that variable over time. (I think that's how it works at least...)
This approach is starting to sound more and more like a genetic algorithm.
Which actually might be a good approach to this problem, since it is
designed for problems where there is no direct route to an optimal
solution and since it is usually fairly good at avoiding getting stuck
in a local minimum ("hill climbing"). And a genetic algorithm will also
(up to a point) get better and better solutions the longer it runs, but
will give you a solution at any time.
- Logan
.
- Follow-Ups:
- Re: Assignmnet problem with rules
- From: Daniel T.
- Re: Assignmnet problem with rules
- References:
- Assignmnet problem with rules
- From: Ayende Rahien
- Re: Assignmnet problem with rules
- From: Willem
- Re: Assignmnet problem with rules
- From: Ayende Rahien
- Re: Assignmnet problem with rules
- From: Daniel T.
- Re: Assignmnet problem with rules
- From: Willem
- Assignmnet problem with rules
- Prev by Date: Re: Help Badly needed- FSA and ASCII stuff
- Next by Date: Re: hex help
- Previous by thread: Re: Assignmnet problem with rules
- Next by thread: Re: Assignmnet problem with rules
- Index(es):
Relevant Pages
|