Neural Networks and Function Approximation

Abstract

In this chapter, the sturctures, approximation properties as well as weights updating algorithms of the neural networks used in this book are overviewed. Firstly, general function approximation abilities of neural networks are briefly introduced. Then, two classes of commonly used neural networks, linearly parameterized networks and non-linearly parametrized networks, are discussed in details respectively. For each class of neural network, after the introduction of network structures, the approximation properties are analyzed. Moreover, the neural network weights learning algorithms are given with the rigorous convergence proof for the easy utilization in the rest part of the book.