Neural Network Representation of Fuzzy Systems

A. VeselĂ˝ (Czech Republic)

Keywords

Fuzzy systems, neurofuzzy systems, neural networks.

Abstract

Fuzzy systems are sometimes represented in a neural network form as so called neuro-fuzzy systems so that the neural network adaptation algorithms might be used for their fine-tuning to a concrete application. A neuro-fuzzy system is a group of mutually connected simple processing units. However, it does not come under paradigm of fuzzy neural network, which is a group of fuzzy neurons connected in a way commonly used in neural network theory. In the article we will show how Mamdani fuzzy systems and zero-order or first-order Sugeno fuzzy systems might be represented with fuzzy neural networks. Fuzzy neural network representations of Sugeno fuzzy systems contain only real neurons. Only a special type of Sugeno fuzzy systems may be represented with a classical neural networks. However, Sugeno fuzzy systems may be with classical neural networks approximated.

Important Links:



Go Back