ABSTRACT
As stated in the previous chapter artificial neural networks are similar to biological networks based on the concept of neurons, connections, and transfer functions. The architectures of various types of networks are more or less similar. Most of the majority variations prows from the various learning rules like Hebb, Hopfield, Kohonen, etc., and the effect of these rules on the network’s topology. This chapter outlines some of the most common artificial neural net-
works based on their major class of application. These categories are not meant to be exclusive, they are merely meant to separate out some of the confusion over network architectures and their best matches to specific applications. The single layer and multi layer networks are discussed in this chapter along with the detailed architecture and algorithm of the prediction networks. The chapter also delineates the basic methodology to implement the prediction networks using MATLAB. Basically, most applications of neural networks fall into the following
four categories:
1. Prediction
2. Classification
3. Data Association
4. Data Conceptualization
Table 3.1 shows the differences between these network categories and shows which of the more common network topologies belong to which primary category. A few of these networks grouped according to their specific practical application are being used to solve numerous types of problems. The feedforward back-propagation network is used to solve
Network Category
Neural Network Applications
Prediction 1. Perceptron 2. Back Propagation 3. Delta Bar Delta 4. Extended Delta Bar Delta 5. Directed Random search 6. Higher Order Neural Networks 7. Self-organizing map into Backpropagation
Used to pick the best stocks in the market, predict weather, identify people with cancer risks etc.
