Hierarchical Connectionist Approach for an Intelligent Modeling System

An innovative strategy for modeling has been developed. A hierarchical structure of a connectionist network is designed to successively handle the complicated scenarios derived from various changing parameters in the real world domains. Inorder to start the intelligent modeling process, some domain knowledge is used to first construct a system taxonomy; then, based on this system taxonomy, a collection of efficient connectionist networks is integrated into a hierarchical framework. Each individual multi- layer perceptron network is trained for a specific task using existing cases in the database. The output of each connectionist network is a set of predicted salient characteristics for simulated results.