Chapter Summary

    Encog is a framework that allows you to create neural networks or bot applications. This book focuses on using Encog to create neural network applications. This book focuses on the overall layout of a neural network. In this chapter, you also saw how to create an Encog application that could learn the XOR operator.

    Neural networks are made up of layers. These layers are connected by synapses. The synapses contain weights that make up the memory of the neural network. Some layers also contain threshold values that also contribute to the memory of the neural network. Together, thresholds and weights make up the long-term memory of the neural network. Context layers will be covered in Chapter 12, Recurrent Neural Networks.

    There are several different layer types supported by Encog. However, these layers fall into three groups, depending on where they are placed in the neural network. The input layer accepts input from the outside. Hidden layers accept data from the input layer for further processing. The output layer takes data, either from the input or final hidden layer, and presents it on to the output layer. The output layer presents the data to the outside world.

    The XOR operator was used as an example for this chapter. The XOR operator is frequently used as a simple “Hello World” application for neural networks. The XOR operator provides a very simple pattern that most neural networks can easily learn. It is important to know how to structure data for a neural network. Neural networks both accept and return an array of floating point numbers.

    This chapter introduced layers and synapses. You saw how they are used to construct a simple neural network. The next chapter will greatly expand on layers and synapses. You will see how to use the various layer and synapse types offered by Encog to construct neural networks.