A Relational Neural Network Database Model

In this paper a framework for the conceptual and physical integration of neural networks into relational database systems is presented. The neural network 'meta model', as the neural network paradigms, the static and dynamic properties, and the pattern description, is realized by a collection of relations and dependencies between them. The specific network objects (the 'ortho' networks) are tuples in these relations. An object oriented approach is followed for description of the paradigm hierarchies, and it is shown, how it can be mapped to the relational model. Further a handling scheme for the application data set, as input, output and training information, stored in the relational database together with the neural networks is depicted. Finally the design of the new NeuroAccess system based on the proposed framework is presented. 1 Introduction Many different approaches for the modeling of neural networks were presented in the literature (e.g. [H + 92]). Generally the object-orie..