Dynamic System Identification Using Pseudo-gaussian-based Recurrent Compensatory Fuzzy Neural Networks

In this paper, a Pseudo-Gaussian-based Recurrent Compensatory Fuzzy Neural Network (PG-RCFNN) is proposed for identification of dynamic systems. The recurrent network is embedded in the PG-RCFNN by adding feedback connections in the second layer, where the feedback units act as memory elements. The compensatorybased fuzzy reasoning method is using adaptive fuzzy operations of fuzzy neural networks that can make the fuzzy logic systems more adaptive and effective. The PseudoGaussian membership function can provide the proposed model, which own a higher flexibility and can approach the optimal result more accurately. An on-line learning algorithm that consists of structure learning and parameter learning is proposed to automatically construct the PG-RCFNN. The structure learning is based on the degree measure and the parameter learning is based on the supervised gradient decent method. Computer simulations have been conducted to illustrate the performance and applicability of the proposed model.