Models of ANNs

During the last fifty years, many different models of artificial neural networks have been developed. A classification of the various models might be rather artificial. However it could be of some benefit to look at the type of data which can be processed by a particular network, and at the type of the training method. Basically we can distinguish between networks processing only binary data, and networks for analog data. We could further discriminate between supervised training methods and unsupervised methods.

Supervised training methods use the output values of a training in order to set up a relationship between input and output of the ANN model. Unsupervised methods try to find the structure in the data on their own. Supervised methods are therefore mostly used for function approximation and classification, while unsupervised methods are most suitable for clustering tasks.