Re: Fuzzines index m in Fuzzy C-Means

Will Dwinnell (76743.1740@CompuServe.COM)
Wed, 8 Jul 1998 06:29:19 +0200 (MET DST)

"Is there any preference for choosing the value of the fuzziness
index m in Fuzzy C-Means. I have seen that many people use the
value m=2 but without giving any reason. They say it can be
chosen arbitrarily in the interval (1, infinite)."

I don't know of any theoretical reason to say that any particular
value is somehow 'optimal'. I can tell you that setting the
fuzziness parameter to 1 causes fuzzy c-means to behanve as the
conventional k-means clustering algorithm and that infinity cause
the cluster centroids to collapse to the centroid of all the
data. My personal experience is that values betwenn 1 and 2 do
well. In a recent "PC AI" article, I believe I used 1.25 or 1.5,
which, given the above, would be "a little fuzzy". With more
than 5 or so dimensions, fuzzy c-means collapsed on data I have
tested when I went much higher than 2.

-- 
Will Dwinnell

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