At 06:01 AM 6/22/2001, ejf wrote:
>I think Chad Loder wrote:
> >
> > "If there exists a diagnosis-fact X and a distinct
> > diagnosis-fact Y s.t. X and Y agree on some particulars
> > but disagree on other particulars, then
> > assert(needs-resolution (X Y))"
> >
> > Is [this] possible within Jess? I see in
> > the documentation that the (exists) CE does not let
> > you bind the triggering fact -- from a performance
> > point-of-view, I understand why that restriction
> > must exist.
>
>Well, the reason for this isn't efficiency -- it's that by definition,
>(exists) is true whether there are one or one hundred matches. There
>is no "one" fact that satisfies an (exists).
>
>But without using (exists), you would get a rule firing for each case
>that matches your criterion, and isn't that what you want, anyway? You
>want to resolve all the conflicts eventually, right? So you want the
>rule to fire for each conflict.
Yes -- I don't know what I was thinking! Something along the lines
of "It can't be possible to compare two separate facts bound in the
same LHS" -- but of course that's not the case.
I just found a paper on "merge logic" which talks about the merge
phase...I came up with some wacky ideas about merging which involve
building a graph of compatible facts and then cutting the graph
into maximal partitions. All the stuff in the Giarratano & Riley
book wrt resolving conflicts seems too complicated.
I don't have any grounding in this field -- I imagine
merging/resolving must be a common problem. Does anyone know of
any standard methodologies for doing this?
Thanks,
c
---------------------------------------------------------------------
To unsubscribe, send the words 'unsubscribe jess-users [EMAIL PROTECTED]'
in the BODY of a message to [EMAIL PROTECTED], NOT to the
list (use your own address!) List problems? Notify [EMAIL PROTECTED]
---------------------------------------------------------------------
