Artificial neural network

From Citizendium
Revision as of 06:36, 29 April 2007 by imported>Derek Harkness (Add workgroup tag and cz live)
Jump to navigation Jump to search

Artificial Neural Network (ANN for short) is a processing model inspired on the biological neural networks. Artificial neural networks are composed by simple nodes called artificial neurons and the processing behavior is stored in the node interconnections as weights.

Adaptation and Learning

When a neuron receives and processes an input signal, it changes its behavior by changing its threshold and/or weight values, modifying the network. Since artificial neurons have a predictable behavior, artificial neural networks can be trained by being fed with sequences of inputs, often determined by certain functions. Besides, there are neural networks which do not require training.