Hierarchical Fuzzy identification using gradient descent and recursive least square method
Zeinab Fallah, Mojtaba Ahmadieh Khanesar, Mohammad Teshnehlab · 2013
In this paper, the parameters of hierarchical fuzzy systems are trained using the simultaneous use of Gradient Descent (GD) for nonlinear parameters and recursive least square (RLS) algorithm for linear parameters. One of the most effective ways to overcome the curse of dimensionality of fuzzy systems is the use of hierarchical fuzzy systems (HFS). Considering the learning abilities of fuzzy systems, two learning algorithms GD and GD+RLS have been used to teach HFS. The results of simulation show that, the use of HFS causes the decrease in the number of rules and results in better performance in identification. In addition, when GD+RLS algorithm is used for learning HFS, it produces better results when it is compared to GD algorithm.