Introduction to Functional Networks

In Chapter 1 we have shown that neural networks are powerful tools to solve a wide variety of practical problems. Neural networks consist of one or several layers of neurons connected by links. Each neuron computes a scalar output from a weighted combination of inputs, coming from the previous layer, using a given scalar activation function (usually step or sigmoidal functions). Since the neural functions are given, the parameters of the neural network are the weights of the connections. Then, learning consists of obtaining the optimal weights to reproduce a given set of data.