The atomic computational unit of a neural network is the artificial neuron. It simulates several basic functions of the biological neuron, evaluates the intensity of each input, sums up the different inputs, and compares the result with an appropriate threshold. Finally, it determines what the output value is. The basic anatomy of the neuron is known and the main biochemical reactions that govern its activity have been identified. A neuron can be considered the elemental computational unit of the brain. In the human brain, about 100 different classes of neurons have been identified. The following figure shows the scheme of a single neuron:
Figure 7.5: Neural network scheme
The main feature of the neuron is to generate an electric potential propagating along the axon (neuron output) when electrical activity at the neuron body level exceeds a certain threshold. Neuron input is a set of fibers called dendrites; they are in contact with the axons of other neurons...