NEUREX-an expert system for the autonomous design of artificial neural nets

Abstract

The learning capacity of artificial neural nets removes the need to formulate a model or to extract knowledge for the implementation of any correspondence function. On the other hand, the design process calls for a certain number of iterations before one can find the right configuration and learning parameters for use in a given application. To mitigate this difficulty, NEUREX has been designed to consist of an expert system and an artificial neural net simulator. It supervises autonomously the process of design in the same way as would an experienced designer, and is currently in readiness to supervise artificial neural net design using standard backpropagation.