Re: CLIPS Question.

Ah... of course.  No need to reset.

thanks

Juliana L Holm wrote:
> Yes. YOu have to test proximity from the point of view of each object.  I'd 
> save the proximity scores in facts, then write a maximizing function of some
> sort to sort them out after all the scoring is done.
>
> But that is just my approach.  
>
> The matching/scoring function would be better done using something akin to Case
> Based reasoning. assuming there are a number of features, but the maximizing
> could be done in a rule base pretty easily, especially if there are rules as
> to how you resolve conflicts. 
>
> Julie
>
> StupendousMan <[EMAIL PROTECTED]> wrote:
> A
>
> > Ok.  Thanks.   So assuming I continue down the path I am on, where 
> > matches with desirable traits are tallied and then reset as a match is 
> > found, is this going to be horribly inefficient?  For each object in 
> > pool 1, I will have to reset and re-run this trait tallying.
>
>
> > Is there maybe a better way just to do that part?
>
>
> > R
>
>
> > Juliana L Holm wrote:
> >
> > > Possibly you could create, using your weighting of desirables a desirability
> > > score and select the highest scoring desirables.
> > >
> > > This may be a job for case based reasoning, not rule based reasoning.  
> > >
> > > That said, I don't know of a free case based shell you could use that would
> > > accomplish this.
> > >
> > > Julie
> > >
> > > StupendousMan <[EMAIL PROTECTED]> wrote:
> > >
> > >
> > > > I think shells is the right group for this question.
> > >
> > >
> > > > I am creating my first clips program and I have had a good read through 
> > > > the Giarratano and Riley book but I think this is more of a taste/style 
> > > > issue.
> > >
> > >
> > > > I basically have two pools of objects that I need to match up based on 
> > > > how well suited they are to each other.  There are numerous ways that 
> > > > objects can match with each other, but in the end I need to end up with 
> > > > one on one pairings of objects.  (no this isn't a dating service :-) )
> > >
> > >
> > > > I have several dozen parameters that the objects could have in common, 
> > > > which can be broken into two groups: mandatory and desirable.  Of course 
> > > > the mandatory parameters must be met, which is easily done in CLIPS, but 
> > > > its the desirable parameters that have me confused.
> > >
> > >
> > > > Naturally, I want the object with the most matches in the desirable 
> > > > group.  I'm really not sure how to accomplish this.
> > >
> > >
> > > > If the desirable traits had some sort of weighting, I could create a 
> > > > rule with every trait, give it the higest salience, then make another 
> > > > rule with slightly lower salience that has one less trait in the LHS.
> > >
> > >
> > > > The only solution I have come up with seems to be very inefficient, and 
> > > > somewhat contrary to the principles of the language, but here goes (in 
> > > > pseudo code)
> > >
> > >
> > > > Add two new slots to the object, matches_mandatory and count_desirables
> > >
> > >
> > > > IF object matches all mandatory constraints
> > > >         THEN set matches_mandatory to true.
> > >
> > >
> > > > IF object matches_mandatory AND matches desirable_quality_1
> > > >         THEN set count_desirables = count_desirables + 1
> > >
> > >
> > > > IF object matches_mandatory AND matches desirable_quality_2
> > > >         THEN set count_desirables = count_desirables + 1
> > >
> > >
> > > > blah ... blah ... blah...
> > >
> > >
> > > > Finally, the object with the highest count is the right one and the rest 
> > > > can be reset for the next object that comes along (since it will have 
> > > > different desirable traits that it is looking for)
> > >
> > >
> > > > Is there a better way?
> > >
> > >
> > > > Rick