Practical Neural Network Recipies in C++

by Timothy Masters

Book Overview

This text serves as a cookbook for neural network solutions to practical problems using C++. It will enable those with moderate programming experience to select a neural network model appropriate to... This description may be from another edition of this product.

• Format: Paperback
• Language: English
• ISBN: 0124790402
• ISBN13: 9780124790407
• Release Date: April 1993
• Publisher: Morgan Kaufmann Publishers
• Length: 493 Pages
• Weight: 2.02 lbs.
• Dimensions: 1.0" x 7.5" x 9.2"

Customer Reviews

The best book to learn Neural Networks

As an undergraduate Math and Computing student, when I took up a high-level course in Neural Networks as an open elective, the only book my instructor recommended was Neural Networks by Simon Haykin. I got that book, and was soon put off by the mathematical rigour it had. It was sometime later that I came across Practical Neural Network Recipes in C++ by Masters'. This, by all standards, is an exceptionally well written book. It has the complete code for a neural network application, including Conjugate Gradient based back-propagation, Simulated Annealing and Genetic Algorithm powered optimisation, and much more. The code, although not very object-oriented, is clear and easy to follow. Undergraduates with a limited knowledge of mathematics will most certainly appreciate the way Masters' deals with the underlying concepts behind neural networks training and use. He simplifies the mathematical equations, and the code listings serve to see the math in action. The more mathematically mature can look into the excellent references provided in the text. When much later in the course I went on to study Recurrent Networks (RNNs, which Masters' doesn't cover in his book), I found myself going back to Masters' when I had to implement algorithms for RNN training. This is one book that will teach you to convert complex mathematical equations into working code. Its a skill that is of much importance to most computational science students. This book is a must have for all neural networks students and practitioners alike.

Very practical indeed

This book is exactly as advertised. Other excellent books on Neural Networks will have you buried in mathematical notation that will challenge even readers with some statistics background and a couple of semesters of Calculus. These books are definitely worth your while if you can handle the math, but even then, translating these books from theories to solving real problems is no easy feat. By contrast, this book presents a good introduction to basic feedforward neural networks that is very readable to users with a moderate math background, and probably readable with some effort for motivated readers with limited math. You can read this book and come away with a reasonable understanding of how a feedforward network functions. Still, that's not even the strength of this book.Not only is this book "practical" in the sense that it is readable, it is practical in that it tackles a host of additional topics necessary for using a neural network in the real world. It discusses annealing and genetic techniques for avoiding local minima. It discusses singular value decomposition for avoiding problems with redundant inputs. It discusses the best ways of building training sets and preparing input data, as well as ways of evaluating the performance of networks and attaching confidence measures. It would be easy to charge right in, use a neural network as a black box, give it a dataset and train it, and then wait for it to pop answers out. The only problem is, this will yield results that are worthless in the real world. All of these concerns have to be addressed to build a model that can actually be used for something.I was very happy with the code base included with this book as well. In addition to a neural network using conjugate gradient descent (as well as Kohonen learning), code is integrated into the main program for annealing, genetic initialization, and singular value decomposition, as discussed in the text. I found the section on how to use the program slightly confusing at first, but once I figured out how to operate it, it was easy to set it up and use it. The code base is C++ that is deeply rooted in C, so it won't impress object-oriented gurus at all, but it should be understandable and fairly easy to work with for users with a good background in C, but who aren't C++ experts. For me, the bottom line is that the code works, it's not hard to understand (in my opinion), and it shouldn't be that hard to extend to perform new functions. In this day and age, it's probably worth mentioning that the program comes with a simple command-line interface, so if you want something that runs in a spiffy GUI, you'll have to write one.I would recommend this book strongly as a first book on neural networks for readers that are interested in learning neural networks in the context of solving practical problems. I would also recommend this book to readers who have a book or two discussing the theoretical aspects of neural networks and want something that will hel

Great intro to neural networks -- Many examples

When I received Pratical Neural Network Recipes in C++, I was pleasantly surprised on how easy it was to follow. Even though I have an extensive calculus background, I believe almost anyone with a background in statistics or college algebra can follow along. As far as content, Masters has shown his ability to explain a complex subject without making it overly complex. I was happy that Masters did not get too in depth with mathmatical proofs. Instead, he sticks to the point -- how to make artificial neural networks that aid in everything from pattern recognition to stock forecasting. He also explains several different kinds of networks such as genetic, hybrid, and multilayer feedforward, and the various benefits and pitfalls of each. The Neural program that comes with the book was also of great help (It's on a 3 1/2 floppy, not a 5 1/4). I recommend this book to anyone who wants to learn about NN's.

Fantasic introduction to neural networks.

The author does a great job with this book. He presents the complex material of neural networks in a very simple manner making it understandable to anyone interested in: (1) finding out more about neural networks, (2) using neural networks in any field, (3) applying neural networks in any field of research (ie: medicine, biology, finance, etc...). The author goes over everything that one needs to know about neural networks -- from the basics to how to implement your own network. Not only does he present the material in a concise manner, but he also gives C++ code to implement a neural network both in the book and on disk. Overall, I think that this is an excellent book to begin with if you are interested in neural networks and their applications.

How to build 'em - How to use 'em - And actual source code!

Easily the best treatment of neural networks I have ever read. Outstanding treatment of the innards, how they work, and years of practical experience boiled down into heuristics for programming (with optimized source code examples!), configuring, training, and evaluating nets. The theory is brilliantly explained within each topical context in lieu of boring chapters on NN theory and math. Mathematical expressions are used only where they add clarity and are not gratuitiously used where the author's excellent English can do the job. And talk about English! Masters is one of those phenoms who speak math and English with equal facility. The writing is simply outstanding. The book is so good it is hard to decide what parts are most valuable. Amazingly, it is as useful for the novice wanting to learn something about neural nets as it is for a professional looking for tips and techniques! I have made the book mandatory reading for my team of knowledge discovery scientists and engineers