Het Categoriseren van Sequenties in een Neuraal Netwerk

In this research the categorization of sequences has been modeled. For this purpose, guidelines have been formulated that need to be followed to model categorization and sequence learning in a biologically plausible way. Literature research states that the connectionist approach applies well to modeling such cognitive processes. Hence, categorizing sequences is modeled in an artificial neural network. The network used in this research is an integration of the Rumelhart feedforward network and the Elman simple recurrent network. To test if the network succeeded in categorizing sequences the network had to learn different routes and categorize them by rotation. Two conclusions follow from the results. First, the network is able to learn the sequences to a certain extent. Second, the network seems able to categorize the learned sequences. The fact that this is an explorative research should be taken into account while reviewing the conclusions.