Original language | English |
---|---|
Pages (from-to) | 755-766 |
Journal | IEEE Transactions on Fuzzy Systems |
Volume | 14 |
Issue number | 6 |
DOIs | |
Publication status | Published (in print/issue) - 1 Dec 2006 |
Bibliographical note
Other Details------------------------------------
This paper proposes a hybrid, genetic algorithm (GA)-based approach to create a fuzzy neural network implementing Takagi-Sugeno fuzzy models. An automatically generated structure is subsequently optimised by a new evolving winner GA based method, where the GA optimises the number of neurons. A hybrid parameter learning approach then adjusts the parameter matrix and the parameters of the membership functions. The importance of the approach is that it can generate a more compact structure than alternative methods, at the expense of being used offline, and is shown to be superior to the OBS (optimal brain surgeon) technique.