ABSTRACT

Details of the elemental building blocks of a neural network, for example, individual neurons, nodal connections, and the transfer functions of nodes, were provided in Chapter 1. Nonetheless, in order to fully understand the operation of the neural network model, knowledge of neuron connectivity and layer arrangement is essential. Connectivity refers to the level of interaction within a system; in neural network terms, it refers to the structure of the weights within the networked system. The selection of the “correct” interaction is a revolving, open-ended issue in neural network design and is by no means a simple task. As will be presented in subsequent chapters, there are countless methods used to aid in this process, including simultaneous weight and structure updating during the training phase and the use of evolutionary strategies: stochastic techniques capable of evolving both the connection scheme and the network weights. Layer arrangement denotes a group of neurons that have specialized function and are largely processed through the system as a collective. The ability to interpret and logically assemble ways in which neurons are interconnected to form the networks or network architectures would thus prove constructive in model development and €nal application. This chapter will provide a solid background for the remainder of this book, especially Chapter 3, given that training of neural networks is discussed in detail.