Abstract
In this paper, the normalized exponential neural network (ENN) is studied. It is proved that ENN is a universal approximator. The stability relation between systems and neural networks working as controllers is investigated. The results show that when designing a system, one should firstly consider system stability rather than controller stability. Accordingly, a new hybrid learning algorithm is presented, and it is proved that this algorithm eventually converge to equilibria.
Similar content being viewed by others
References
Wang L Xet al. Fuzzy basis functions, universal approximation and OLS.IEEE Neural Networks, 1992.
Hirsch M W. Convergent activation dynamics in continuous time networks.Neural Networks, 1989, (2).
Moody Jet al. Fast learning in networks of locally-tuned processing units.Neural Computations, 1989 (1).
Hirsch M Wet al. Differential Equations, Dynamical Systems and Linear Algebra. Academic Press, New York, 1974.
Hertz Jet al. Introduction to the Theory of Neural Netowrks. Addison-Wesley Press, 1991.
Patyera M Jet al. Fuzzy Logic. Willey Teurner Press, 1996.
Beneveniste Aet al. Algorithm Adaptatifs et Approximations Stochastiques. Masson, Paris, 1987.
Lee M Het al. Neuro-fuzzy identifiers and controller for fuzzy systems. InProc. of IFSA'93.
Gupta M Met al. On principles of fuzzy neural networks.J. Fuzzy Sets and Systems, 1994.
Kosko B. Neural Networks and Fuzzy Systems. Prentice-Hall, 1992.
Benaim M. On functional approximation with normalized Gaussian units.Neural Computations, 1994.
Wang Shitonget al. Apply the new hybrid algorithm to improve the Ishibuchi's method of fuzzy if-then rules with multilayer feed-forward neural networks. InProc. of ICICS'96.
Wang Shitonget al. The improved learning method for Ishibuchi's method of fuzzy if-then rules with multilayer feedforward network. InProc. of FUZZ-IEEE'96.
Wang Shitonget al. Fuzzy inductive learning algorithm ILA*.Int'l J. Fuzzy Sets and Systems (have been accepted).
Author information
Authors and Affiliations
Additional information
Wang Shitong is a Professor and the head of the Department of Computer at East China Shipbuilding Institute. Up to now, he has published 5 books and 92 papers in international/national journals or in international conferences. His research interests include artificial intelligence, neural networks, fuzzy system and rough set theory.
Rights and permissions
About this article
Cite this article
Wang, S. Normalized exponential neural networks. J. of Comput. Sci. & Technol. 13, 375–383 (1998). https://doi.org/10.1007/BF02946626
Received:
Revised:
Issue Date:
DOI: https://doi.org/10.1007/BF02946626