Neural networks are applied mainly where classic computers fail. The problems for neural networks do not have known algorithm, or analytical description it too complicated for computer processing. Their main application field is for problems with large example data that cover sufficient area of the problem. The basic application fields of neural networks are described in following: Economic information systems, Technology and manufacture, Health care, Meteorology.
Artificial neural networks have some advantageous properties. They can implement an arbitrary transformation above input date, so they are universal. A neuro-computer based on neural network does not need to be programmed, as its required behavior can be achieved by training with appropriate examples. Neural networks are, thanks to a large number of neurons and connections (synapses) and thanks to the fact that the information is distributed over the entire network, very robust. Failures of neurons lead only to a slow degradation of the network. They have an ability of generalization, abstraction, i.e. to react equally on a certain set of input data, not only to certain element from this set.