The intention of this book is to bring together wavelet development and common learning paradigms in order to propose more powerful methods for classification, prediction, and pattern recognition. It has to be regarded as a first attempt to unite these topics in a single volume. Since the book does not explore these topics in any great depth, readers are not expected to be highly proficient in either area, although it would certainly help.
A large chapter on basic mathematical material begins the book, giving a useful background for the rest of the content. This first chapter is more of a review, since statements and theorems are given with neither proofs, nor justifications. It is oriented more towards functional analysis. Next, there is a surprisingly short chapter on wavelets. At just 20 pages, this is not much for an uninitiated reader. Neural networks are given a more thorough treatment in the next chapter, although again it remains cursory (for example, a page and a half on backpropagation). The merging of the wavelet basis functions with neural network topology towards a powerful universal function approximation mechanism is the topic of chapter four, and gives the book its flavor of originality. Readers acquainted with wavelets and/or neural networks might prefer to jump directly to this chapter.
Original contributions from several authors are contained in the second part, which is oriented more toward applications. Standing high above the commonplace are the chapters on predicting chaotic series, concept learning, and separating order from disorder. The two remaining chapters, “Recurrent Learning” and “Radial Wavelet Networks,” are interesting enough, but the subject matter is more ordinary.
A well-structured bibliography is also included. Overall, as a first attempt to present ideas from these fields merged together, the book is convincing. However, it is still a long way away from the practical style of writing needed to bridge the gap between theoretical studies and hints for applications, and the level where algorithms are structured and implemented.