Essentials of artificial neural networks

Abstract

Artificial neural networks have been studied for more than five decades since the pioneering work of McCulloch and Pitts [193] in which they proposed a model of an artificial neuron, and slightly later Hebb’s psychological study [133] pointing out the importance of the connections between artificial neurons to the process of learning. Artificial neural networks are a new generation of biologically-inspired, massively-parallel, distributed information processing systems. They consist of processing elements (also called nodes, units or artificial neurons) and connections between them with coefficients (weights) bound to these connections, which constitute the neuronal structure, as well as learning algorithms attached to this structure. They are also called connectionist systems because of the main role of the connections in them; the connection weights play the role of the “memory” of the system.