Re: AI

WSiler (wsiler@aol.com)
Sun, 27 Sep 1998 00:45:19 +0200 (MET DST)

>Thus, re-entrant and recursive rules are just one problem with the current
>family of fuzzy shells.

Amen, brother.

Perhaps it would be worth while to list the facilities which an ideal fuzzy
system shell should have. Of course, none of us has an ideal shell, but among
those who have actually worked with real-world fuzzy and non-fuzzy expert
systems it might be possible to work out a consensus so at least we have an
idea as to where we ought to be going. Here are a few off-hand thoughts, partly
derived from your work, partly from Bill Combs' work, and partly from our own
experience.

1. It should be possible for rules to be re-entrant and recursive. (We define
recursivity as the ability of a rule or block or rules to make themselves
refireable until spme goal is reached.

2. Redefine fuzzy numbers to eliminate the convexity requirement, and seriously
investigate what can be done when fuzzy numbers are considered as fuzzy sets,
and hence subject to set-theoretic operations.

3. Seriously investigate the strengths and possible limitations of Combs' Union
Rule Configuration model.

4. Check the limitations of approximate reasoning methods, and consider direct
assignment of consequent variable values as an alternative.

5. Seriously investigate the default initial truth value of variables whose
value has not yet been established. This relates to approximate reasoning.

6. Seriously investigate the way in which memory modification by rules of
variables which have an established value should be handled.

7. Consider the uses of both parallel and sequential rule-firing schemes.

8. Consider possible variations of schemes for firing successive rule blocks in
multistep reasoning.

9. Try to establish what data types should be included in fuzzy expert systems,
and how many levels of fuzziness to allow.

10. Consider under what circumstances the laws of excluded middle and
contradiction should be obeyed and when they should be ignored.

11. Consider the strengths of conventional AI, and which of its features should
be incorporated in fuzzy expert systems.

All this is right off the top of my head. Lord only knows what I could come up
with if I took time to think about it.

Gosh, Earl, are we writing to anyone except ourselves?

William Siler

Odium ignorantem est odium infantem,
sed odium savantem est odium ferentem!

############################################################################
This message was posted through the fuzzy mailing list.
(1) To subscribe to this mailing list, send a message body of
"SUB FUZZY-MAIL myFirstName mySurname" to listproc@dbai.tuwien.ac.at
(2) To unsubscribe from this mailing list, send a message body of
"UNSUB FUZZY-MAIL" or "UNSUB FUZZY-MAIL yoursubscription@email.address.com"
to listproc@dbai.tuwien.ac.at
(3) To reach the human who maintains the list, send mail to
fuzzy-owner@dbai.tuwien.ac.at
(4) WWW access and other information on Fuzzy Sets and Logic see
http://www.dbai.tuwien.ac.at/ftp/mlowner/fuzzy-mail.info
(5) WWW archive: http://www.dbai.tuwien.ac.at/marchives/fuzzy-mail/index.html