GD+FC learning algorithm for system modeling
Yonghung Tan, Chun‐Yi Su, Xuanju Dang · 2002
A gradient descent plus fuzzy control (GD+FC) learning strategy is proposed. In this method, the learning procedure is considered as a feedback control system that consists of a controlled process, a feedback mechanism, and a feedback controller. Therefore, the fuzzy control technique may be implemented in order to achieve fast and stable convergence in the learning procedure. After that the convergence feature of the proposed learning algorithm is investigated. Then, the proposed algorithm is used to train neural networks for system modeling. A comparison of the proposed algorithm with the other learning approaches, e.g. GD and PIDGD methods, is also illustrated. Finally, the article presents an example of system modeling for a temperature process with the proposed learning approach.