The formal neural networks represent a new view on ICT. The original neural network was based on the work by Warren McCulloch and Walter Pitts published in 1943. The most important issue in investigating neural networks is to consider how they may learn. Various kinds of artificial neuron based on the McCulloch-Pitts unit will now be described, and several learning rules discussed. The Linear Threshold Unit We have discussed in the classical neuron and the simple, on/off threshold unit. As McCulloch-Pitts had demonstrated in the early 1940s, to a first approximation, networks of such units can perform arbitrary computations.