Re: Looking for Pattern Matching and Learning method

Randolph M. Jones (rjones@colby.edu)
Wed, 28 Apr 1999 03:41:35 +0200 (MET DST)

I think you would be interested in approaches that do "conceptual
clustering"
or "unsupervised learning". Doug Fisher at Vanderbilt
(http://www.vuse.vanderbilt.edu/~dfisher/) has done some of the top work
in this area, beginning with his dissertation work on a system called
COBWEB. Another system, called AUTOCLASS and developed at NASA, has
also
had success solving problems of this type. I haven't watched the
literature
closely in recent years, so I'm sure there are also other systems you
cold look at.

Randy Jones

Matt Sullivan wrote:
>
> Hello all,
>
> I am looking for a pattern matching and learning algorithm that will allow
> me to do the following. I have a list of concepts (known or learned by the
> system), specified by a person, that attempts to describe some domain that
> is formally specified within the system. Each concept can have a list of
> attributes associated with it. The problem is, how to take the list of
> concepts and attempt to match them to the domain that is most applicable
> within the system, and to learn from mistakes made.
>
> For example, say a person entered the following concepts (the syntax, made
> up just to describe this problem, is <concept, attributes>): <'alloy
> wheels', yes>, <'engine': 2 litre>, <'color', red>, <'milage', 10,000>. The
> system has a large collection of domains, including one for a car, which
> internally has a list of concepts such as make, model, colur, engine, etc.
> There is also a description of motorbike, with much the same concepts.
> However, the concepts within these domains may have constraints, so , for
> example, the engine concept in motorbike would indicate that a 2 litre
> engine is unlikely, whereas it is fine for a car. Note that I am not
> attempting to match a user description to a particular car, just to the car
> domain.
>
> Can anyone point me to any algorithms or methods that could perform these
> sorts of tasks?
>
> Thanks,
>
> Matt Sullivan
>
> matt@europe.com

############################################################################
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