Neural computation with ideal approximation elements

We describe an analytical framework for the adaptations of neural systems t hat adapt their internal structure on the basis of subjective probabilities constructed via computation of randomly received input signals. A principl ed approach is provided that has the key property that it defines a probabi lity density model that allows to study the convergence of the adaptation p rocess. Certain neural network models (e. g. topological feature maps and a ssociative networks) can be derived from our approach.