Choice Review

Rao and Rao combine a brief survey of the rapidly growing field of neural networks with mostly simple example programs presenting many different types of networks proposed and used successfully. The programs, whose presentation takes up more than half the book (and are supplied on the accompanying disk) are written in the object-oriented C++ language, but could be adapted to other languages. The presentation of the neural network material is not always complete or easy to follow--the authors' English is sometimes awkward and unclear. A student might be better advised to begin a study of neural networks with books such as Murray Smith's Neural Networks for Statistical Modelling (1993), or Bart Kosko's Neural Networks and Fuzzy Systems (CH, Dec'91), then use the Raos's work as a source of programs for experimentation. The programs cover a wide variety of network types. The back-propagation model presented is more elaborate and flexible, and is applied to predicting financial markets. One model briefly presents the idea of "fuzzy" sytems. A useful supplement for projects on neural networks. Advanced undergraduate through faculty. H. D. Warner; Western New England College