Artificial Neural Networks

Abstract

Artificial neural networks (ANN) enable problem solving by changing the structure of interconnected components. In analogy to the biological model, these interconnected elements are referred to as neurons and make up the basic units of information processing. A simple model of a biological neuron is composed of the following components, illustrated in figure 9.1: A large number of short nerve fibers called dendrites conduct impulses from the adjacent neurons to the cell body. The cell reacts to this input with an impulse that is carried to other cells by an axon. An axon is a long nerve fiber that may be up to one meter in length and undergoes extensive branching at its end which enables communication with dendrites or other cell bodies. The neurons are connected via synapses. When an electrical impulse runs through an axon, this causes the synapse to release a substance called a neurotransmitter which is picked up by a neighboring dendrite and changes the electric potential of the cell. When a particular threshold value is reached, an action potential is generated in a cell that in turn is sent along the axon through the synapses to the dendrites of the connected neurons. The human brain has approximately 10¹¹ interconnected neurons whose reaction time is estimated at one millisecond. The average processing time for the recognition of complex visual and/or acoustic information by a human brain is approximately 0.1 seconds. If biological neuronal systems would work like strictly sequential digital computers, only 100 instructions would be possible in that time frame [12, p. 172].