On the significance of internal representations in neural networks

Summary form only given. For a straightforward materialization of internal representations, semantic networks are suggested. In their original form they comprise a graph structure with nodes and links. The nodes may stand for items or concepts (sets of attributes), whereas the links usually indicate relations. In view of the contemporary neurophysiological data, such a degree of specificity and spatial resolution is highly improbable in biology. It must be realized that in the semantic-network model of the brain, predisposition of semantic meaning to the nodes and links has to be postulated. Forms of hardware simulation are considered, and it is shown that self-organizing maps are able to represent concrete and abstract features present in sensory data, as well as syntactic and semantic relations that appear in simple linguistic expressions