Artificial neural networks are a narrow-sensed abstraction of the human brain, thus the organization of the artificial neural system is very similar to the one of biological neurons. The comprehensive understanding of biological neurons is not complete; however, the basic functionality that contributes to the learning ability of a system is implemented in artificial neural networks. The fundamental element, an artificial neuron, is a model based on known behavior of biological neurons that exhibit most of the characteristics of human brains that we are interested in [Vel98]. This is the most significant difference from conventional computers, which have internal fixed instructions to perform specific functions.
Artificial neural networks can be also described as highly parallel distributed computing models. The fundamental processing units, neurons, are highly connected with strengths, which are dynamically changed during the system's learning process.
The discussion in following sections approach the engineer's perspective of understanding the artificial neural networks. Though artificial neural networks are not an exact copy of biological human brain, it is important to begin with understanding fundamental concepts of biological neurons and the human brain.