Neuro-fuzzy systems and their applications
Bogdan M. Wilamowski · 2002
Computational intelligence combines neural networks, fuzzy systems and evolutional computing. Neural networks and fuzzy systems have already proved their usefulness and have been found useful for many practical applications. We are at the beginning of the third technological revolution. Now neural networks are being capable of replacing highly skilled people with all their experience. The concept of artificial neural networks is presented, underlining their unique features and limitations. A review and comparison of various supervised and unsupervised learning algorithms follows. Several special, easy to train, architectures are shown. The neural network presentation is illustrated with many practical applications such as speaker identification, sound recognition of various equipment as a diagnosis tool, written character recognition, data compression using pulse coupled neural networks, time series prediction, etc. In the latter part of the presentation, the concept of fuzzy systems, including the conventional Zadeh approach and Takagi-Sugano architecture, is presented. The basic building blocks of fuzzy systems are discussed. Comparisons of fuzzy and neural systems are given and illustrated with several applications. The fuzzy system presentation is concluded with a description of the fabricated VLSI fuzzy chip.