ANNs have been around almost as long as computers have, and indeed were originally constructed out of electrical hardware. One of the first ANNs was developed in the 1970s to adaptively filter echoes out of phone line transmissions. Despite their initial early success, ANNs waned in popularity until the mid-1980s, when the backpropagation training algorithm was popularized.
ANNs are modeled on our understanding of biological brains. An ANN contains many neurons that connect to one another. The manner, structure, and organization of these neuronal connections is called the topology (or shape) of the network. Each individual neuron is a simple construct: it accepts several numerical input values and outputs a single numerical value, which may in turn be transmitted to several other neurons. The following is a simple, conceptual example of a...