On the mapping capabilities of fuzzy inference systems
Chi-Cheng Jou · 2003
It is demonstrated that fuzzy inference systems can be used effectively for function approximation. A class of fuzzy inference systems under consideration has the characteristic that the output fuzzy set of every fuzzy logical rule is a singleton. It is shown that the class of fuzzy inference systems suggested is capable of approximating any real continuous function to any desired degree of accuracy, provided sufficient fuzzy logical rules are available. To investigate a direct application of results, the relationships between these fuzzy inference systems and multilayer neural networks are discussed.>