Title :
FLIP-net: a network representation of fuzzy inference procedure and its application to fuzzy rule structure analysis
Author :
Maeda, Akira ; Ichimori, Toshihide ; Funabashi, Motohisa
Author_Institution :
Hitachi Ltd., Kawasaki, Japan
Abstract :
A method for analyzing dynamic behavior of fuzzy rules using FLIP-net (fuzzy logic inference procedure network), a network representation of a fuzzy inference procedure, is proposed. Using the FLIP-net representation, errors of fuzzy inference output can be backpropagated. Since backpropagated error indicates how each component of fuzzy logic contributes to the output error, analysis of the error distribution over FLIP-net gives useful information about the dynamic behavior of fuzzy rules. It is shown that fuzzy rule deficiencies, such as redundant, missing, or contradictory fuzzy rules, can be detected by this method, and thus efficient addition and/or modification of fuzzy rules becomes possible
Keywords :
backpropagation; error analysis; fuzzy logic; inference mechanisms; uncertainty handling; FLIP-net; backpropagated error; dynamic behavior; error distribution; fuzzy logic inference procedure network; fuzzy rule structure analysis; output error; Backpropagation algorithms; Computer networks; Error analysis; Fuzzy logic; Fuzzy neural networks; Fuzzy systems; Inference algorithms; Information analysis; Laboratories; Supervised learning;
Conference_Titel :
Fuzzy Systems, 1993., Second IEEE International Conference on
Conference_Location :
San Francisco, CA
Print_ISBN :
0-7803-0614-7
DOI :
10.1109/FUZZY.1993.327430
