DocumentCode :
428588
Title :
Type-2 fuzzy activation function for multilayer feedforward neural networks
Author :
Karaköse, Mehmet ; Akin, Erhan
Author_Institution :
Dept. of Comput. Eng., Firat Univ., Elazig, Turkey
Volume :
4
fYear :
2004
fDate :
10-13 Oct. 2004
Firstpage :
3762
Abstract :
This paper presents a new type-2 fuzzy based activation function for multilayer feedforward neural networks. Instead of other activation functions, the proposed approach uses a type-2 fuzzy set to accelerate backpropagation learning and reduce number of neurons in the complex net. Furthermore, the type-2 fuzzy based activation function provides to minimize the effects of uncertainties on the neural network. Performance of the type-2 fuzzy activation function is demonstrated by exor and speed estimation of induction motor problems in simulations. The comparison among the proposed activation function and commonly used activation functions shows accelerated convergence and eliminated uncertainties with the proposed method. The simulation results showed that the proposed method is more suitable to complex systems.
Keywords :
backpropagation; feedforward neural nets; fuzzy set theory; multilayer perceptrons; transfer functions; backpropagation learning; complex net; complex systems; induction motor problems; multilayer feedforward neural networks; speed estimation; type-2 fuzzy activation function; Acceleration; Backpropagation; Feedforward neural networks; Fuzzy neural networks; Fuzzy sets; Induction motors; Multi-layer neural network; Neural networks; Neurons; Uncertainty;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Systems, Man and Cybernetics, 2004 IEEE International Conference on
ISSN :
1062-922X
Print_ISBN :
0-7803-8566-7
Type :
conf
DOI :
10.1109/ICSMC.2004.1400930
Filename :
1400930
Link To Document :
بازگشت