Neural networks, or better artificial neural networks (ANN) are a family of algorithms especially usefull for classification and function approximation. Their special importance lies in the fact that they provide a versatile way to represent a general nonlinear mapping between multidimensional spaces.

There are lots of different types of artificial neural networks and training methods to optimize their parameters. One common way to categorize artificial neural networks is the distinction between supervised and unsupervised learning. In supervised learning the mapping of some input to some known ouput is learnt and the better the output is predicted the better the ANN performs. In unsupervised learning, on the other hand, the network relies entirely on the input data without reference to any ouptut data. Here, the goal is normally to model the probability distribution of the data or to discover clusters or other structure.

Applications of unsupervised as well as supervised methods to the analysis of biological activity data are presented. Among the multitude of available methods Self-Organizing Feature Maps (Kohonen Networks) and Multi-layer Perceptrons are discussed in depth.

Back to Workshop Schedule

1996-1997 Mathematics in High Performance Computing