Pattern Recognition Using Neural Networks: Theory and Algorithms for Engineers and Scientists

Rating:
Summary: If you want to learn about neural nets DO NOT BUY THIS BOOK!
Review: I started writing out all the things I hated about this book and toped out the 1000-word limit, so I decided to just keep it short.

The first thing I hated about this book is that, while it advertises itself as being a general resource for pattern recognition using neural network technology, it covers only a few aspects of neural networks. The book provides many learning algorithms that can be applied to the multi-layered perceptron, back-propagation, radial-basis, and hopfield networks. Few other network types are used and none of them are described in very good detail.

It does provide a plethora of learning algorithms; most of which apply to the MLP. However, most of these algorithms are so obscure and task-specific they are not likely to help you in your application development.

Few mathematical proofs or derivations of the learning algorithms are given, so you have to leave it to faith that they will work the way you expect them to.

The text is difficult to follow and rarely gives you all the information necessary to understand the concept behind the learning algorithms. Very often, you are forced to examine the pseudo code to understand what the algorithm is really doing.

The Pseudo code for each algorithm cannot be easily translated to a real programming language and it often contains bugs and severe algorithm inefficiencies.

The book tends to focus on supervised learning strategies and barely dips into the huge field of unsupervised learning. Important unsupervised-based neural networks such as the counter-propagation, adaptive resonance theory, and self-organizing map networks are not even mentioned.

If you want to learn about neural networks, BUY A DIFFERENT BOOK! "Neural Networks: A Comprehensive Foundation" by Simon Haykin and "Neural Networks: Algorithms, Applications, and Programming Techniques" by Freeman and Skapura are much better texts on the subject.

Rating:
Summary: A very good reference book in Neural Networks
Review: In this book, key concepts of neural networks, fuzzy logics are introduced through detailed algorithms and graphical illustrations. The book includes some topics not covered or not emphasized in other similar books. This book is written in a clear and expository style with extensive mathamatical derivations, which is well suited to the graduate level students and AI researchers.

I definitely think this is a good book.

Rating:
Summary: A good book for a researcher
Review: This is not a book for a undergraduate student but a good book for a serious researcher. Some of the topics are discussed deeply and extensively, such as the relationship between pattern recognition and neural networks. It looks like that the drawback of this book is that some of the topics are not explained plainly and need to read other reference book to understand them.

Rating:
Summary: I wish I knew about this book earlier!
Review: This is simply the best introduction in print to the most useful types of neural networks that engineers use. Engineering seniors and graduate students should benefit greatly. I was quite impressed at the full-propagation algorithm that is 40% faster than using epochs with backpropagation that tend to thrash (learn/unlearn as it trains). I found the book loaded with ideas and references and have put some of them to good use so far. This was the book I really needed but never knew about until my professor told me to read the section on radial basis function NNs. I put the pseudo-code directly into a C program and it worked beautifully on my data.