Figure 2.

In the left diagram, McCulloch-Pitts neurons x₁ and x₂ each send binary activity to neuron y using the rule y = 1 if x₁ + x₂ > 1 and y = 0 otherwise; this corresponds to the logical AND operator; other logical operators NOT, OR, NOR may be similarly implemented by thresholding. In the right diagram, the general form of output is based on thresholding linear combinations, i.e., y =1 when ∑w_ix_i >c and y = 0 otherwise. The values w_i are called synaptic weights. However, because networks of perceptrons (and their more modern artificial neural network descendents) are far simpler than networks in the brain, each artificial neuron corresponds conceptually not to an individual neuron in the brain but, instead, to large collections of neurons in the brain.