Computational Intelligence: Synergies of Fuzzy Logic, Neural Networks and Evolutionary Computing

5

Neural Systems and Applications

5.1 Introduction

Neural networks (NNs) are meant to interact with the natural environment, and information about the latter is usually collected from the real world through very noisy but redundant sensory signals. On the other hand, in the control of effectors or actuators, one often has to coordinate many mutually dependent and redundant signals. In both cases, neural networks can be used to implement a great number of implicitly and/or poorly defined transformations between variables (Rosenblatt, 1958; Minsky and Papert, 1959; Widrow and Hoff, 1960; Werbos, 1974; McClelland and Rumelhart, 1986). Many applications have been found in simple mathematical methods, such as fitting nonlinear functional expansions into experimental data, pattern recognition and data clustering. The 1988 DARPA Neural Network Study lists various applications of NNs, beginning in about 1984 with an adaptive channel equalizer. This device, which is an outstanding commercial success, is a single neuron network used in long-distance telephone systems to stabilize voice signals. NNs have been applied in many other fields since the DARPA report. Some of the applications mentioned in the literature are as follows:

• Aerospace – high-performance autopilot, flight path simulation, aircraft control systems, fault detection system.
• Automotive – automobile automatic guidance system.
• Banking, Financial and Business – cheque and other document reading, credit application evaluation, ...