Book Image

Neural Network Programming with Java

By : Alan M. F. Souza, Fabio M. Soares
Book Image

Neural Network Programming with Java

By: Alan M. F. Souza, Fabio M. Soares

Overview of this book

<p>Vast quantities of data are produced every second. In this context, neural networks become a powerful technique to extract useful knowledge from large amounts of raw, seemingly unrelated data. One of the most preferred languages for neural network programming is Java as it is easier to write code using it, and most of the most popular neural network packages around already exist for Java. This makes it a versatile programming language for neural networks.</p> <p>This book gives you a complete walkthrough of the process of developing basic to advanced practical examples based on neural networks with Java.</p> <p>You will first learn the basics of neural networks and their process of learning. We then focus on what Perceptrons are and their features. Next, you will implement self-organizing maps using the concepts you’ve learned. Furthermore, you will learn about some of the applications that are presented in this book such as weather forecasting, disease diagnosis, customer profiling, and characters recognition (OCR). Finally, you will learn methods to optimize and adapt neural networks in real time.</p> <p>All the examples generated in the book are provided in the form of illustrative source code, which merges object-oriented programming (OOP) concepts and neural network features to enhance your learning experience.</p>
Table of Contents (19 chapters)
Neural Network Programming with Java
Credits
About the Authors
About the Reviewer
www.PacktPub.com
Preface
Index

Common issues in neural network implementations


When developing a neural network application, it is quite common to face problems regarding how accurate the results are. The source of these problems can be various:

  • bad input selection

  • noisy data

  • very big dataset

  • unsuitable structure

  • inadequate number of hidden neurons

  • inadequate learning rate

  • insufficient stop condition; and/or

  • bad dataset segmentation

The design of a neural network application sometimes requires a lot of patience and trial-and-error methods. There is no methodology stating specifically the number of hidden units and/or which architecture should be used, but there are recommendations on how to properly choose these parameters. Another issue that programmers may face is a long training time, which often causes the neural network to not learn the data. No matter how long the training runs, the neural network won't converge.

Tip

Designing a neural network requires the programmer or designer to test and redesign the neural structure as...