"ANFIS: Adaptive Network Based FIS", Roger Jang,
IEEE Trans. on Systems, Man and Cybernetics, Vol 23, No 3, 1993.
 "Neuro Fuzzy and Soft Computing", Jang Sun Mizutani., Prentice
I am using ANFISEDIT in Fuzzy Logic Toolbox (MATLAB 5.2), and found it
very useful for my research.
I have written code in C++ with the help of reference . It is
working properly, but when I compared it with ANFISEDIT the error
given by my code is more. As given in , I have used hybrid rule
where steepest gradient is used in backward pass and LSE in forward
So why is my code is showing more error than your ANFISEDIT under same
conditions.? What are the improvements that are needed in ?
My project requires use GARIC architecture[Berenji an Khedkar], where
desired output is not available but the reinforcement is available
from Action Evaluation Network(AEN). On page no. 483 of  I found
that ANFIS is similar to Action Selection Network(ASN) of GARIC, but
the problem is, in that case, Hybrid Learning Rule can not be used as
desired output is not available. As we know, steepest gradient method
is slow and may get trapped in local minima. * This severely limits
the use of Hybrid Learnig rule as far Reinforcement Learning is
considered *. So what is the solution to improve the learnig in FIS
when desired output is not available and reinforcemnt(penalty) is
available?. Please provide me some information on this issue.
Thanks. With Best Regards.
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 firstname.lastname@example.org
(2) To unsubscribe from this mailing list, send a message body of
"UNSUB FUZZY-MAIL" or "UNSUB FUZZY-MAIL email@example.com"
(3) To reach the human who maintains the list, send mail to
(4) WWW access and other information on Fuzzy Sets and Logic see
(5) WWW archive: http://www.dbai.tuwien.ac.at/marchives/fuzzy-mail/index.html
This archive was generated by hypermail 2b30 : Mon Oct 08 2001 - 15:32:02 MET DST