A recurrent neuro-fuzzy network structure and learning procedure

A novel recurrent neurofuzzy network is proposed in this paper. This model is constructed from fuzzy set models of neurons. The network has a multilayer, recurrent structure whose units are modeled through triangular norms and conorms, and weights are defined within the unit interval. The learning procedure developed is based on two main paradigms, the gradient search and associative reinforcement learning, that is, the output layer weights are adjusted via an error gradient method whereas a reward and punishment scheme updates the hidden layer weights. The recurrent neuro-fuzzy network is used to develop a model of a nonlinear process. Numerical results show that the neuro-fuzzy network proposed provides an accurate process model after a short period of learning time.

[1]  Witold Pedrycz,et al.  Fuzzy Set Based Neural Networks: Structure, Learning and Application , 1999, J. Adv. Comput. Intell. Intell. Informatics.

[2]  Geoffrey E. Hinton,et al.  Learning representations by back-propagating errors , 1986, Nature.

[3]  Witold Pedrycz,et al.  Fuzzy-set based models of neurons and knowledge-based networks , 1993, IEEE Trans. Fuzzy Syst..

[4]  Hideo Tanaka,et al.  Trainable fuzzy classification systems based on fuzzy if-then rules , 1994, Proceedings of 1994 IEEE 3rd International Fuzzy Systems Conference.

[5]  Ching-Hung Lee,et al.  Identification and control of dynamic systems using recurrent fuzzy neural networks , 2000, IEEE Trans. Fuzzy Syst..

[6]  Kwang Y. Lee,et al.  Diagonal recurrent neural networks for dynamic systems control , 1995, IEEE Trans. Neural Networks.

[7]  Kumpati S. Narendra,et al.  Identification and control of dynamical systems using neural networks , 1990, IEEE Trans. Neural Networks.