Overfitting in Fuzzy System Identification


Subject: Overfitting in Fuzzy System Identification
From: Ali Ghodsi Boushehri (ali@cs.concordia.ca)
Date: Wed Aug 16 2000 - 13:48:29 MET DST


In experiment with subtractive clustering based fuzzy system
identification method I observed that, a model does not have quite as
good performance on the testing data (new data) as on the training data.

The smaller values of training data error do not always follow the
smaller value of testing data error. As a matter of fact, up to a
certain point, the model error for the testing data set tends to
decrease as the training error decreases. After that, attempts to
decrease training error make an unexpected increase in error on testing
data error. What is the justification behind this observation? Is this a
phenomena similar to overfitting in neural networks or there is another
reason for such observation? I would appreciate any comment, paper, etc.
in this regard.

Best regards,

Ali Ghodsi

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



This archive was generated by hypermail 2b25 : Wed Aug 16 2000 - 14:11:56 MET DST