Item

Abstract

This book treats wavelet networks which unify universal approximation features of neuronal networks and multiresolution features of wavelet basis. It is introduced by the authors as the successor of Iyengars famous book Wavelet Analysis with Applications to Image Processing.

The material in the book is organized into two parts. Part I consists of four chapters which cover mathematical preliminaries, Wavelets, Neural Networks and Wavelet Networks. After stressing basic properties of wavelets and neural networks the idea of wavelet networks is introduced.

Part II deals with potential applications and consists of five chapters Recurrent Learning, Separating Order from Disorder, Radial Wavelet Neural Networks, Predicting Chaotic Time Series and Concept Learning. The chapters on Radial Wavelet Neural Networks and Predicting Chaotic Time Series are contributed from guest authors.

The overall structure of this book is more likely a collection of detailed but independent papers, providing the reader exercises at the end of each chapter. Though in comparison to the mathematical precision and systematic treatment of its excellent predecessor, the first part of this book is sometimes too superficial to be informative. The second part is somewhat better—and the chapter on Predicting Chaotic Time Series actually contains the detailed description of an algorithm and provides some practical advice on choosing parameters and network structure. Other practical issues like generalization, validation and overfitting, however, are rarely if at all treated. The reproducibility criterion is not met.

Furthermore, this book suffers from very poor editing. Partially not recognizable illustrations, new sections starting in the middle of a sentence and figures not referenced in the text leave the reader dissatisfied.