The theoretical roots of the approach of an artificial simulation oy means of neural networks lie back in the 1940s. Meanwhile, the claim of modelling biological dynamics has been reduced to a more pragmatic application of the great abilities of net-based concepts and tools. Today, two approaches are mainly in use. They can complement each other, in particular in decision making processes that occur in team sports. Unsupervised or self-organizing maps or networks. (SaM) can learn and recognize the patterns of match situations on their own, whereas supervised or feed forward networks (FFN) can learn by supervision what are the best solution for a situation, once recognized. An FFN consists of a number of layers, each of which contains a number of neurons. In this chapter, the focus is on pattern recognition and SOMs. Supporting games by means of decision-optimizing FFNs has been done with simulated games, e.g. in the field of robot soccer, but currently is still too complicated for original games played by humans.
© Copyright 2014 Complex systems in Sport. Veröffentlicht von Routledge. Alle Rechte vorbehalten.