Neural Network Toolbox

Adaptive Linear Network Architecture

The ADALINE network shown below has one layer of S neurons connected to R inputs through a matrix of weights W.

This network is sometimes called a MADALINE for Many ADALINES. Note that the figure on the right defines an S-length output vector a.

The Widrow-Hoff rule can only train single-layer linear networks. This is not much of a disadvantage, however, as single-layer linear networks are just as capable as multilayer linear networks. For every multilayer linear network, there is an equivalent single-layer linear network.

Linear Neuron Model

Single ADALINE (newlin)