This is the first-year funding of a three-year continuing award IRI-9214141. Both rule-based and neural-network approaches have been useful for the development of intelligent systems. The main objective of this research is to develop a robust machine learning and inference system by combining these two approaches. A novel prototype system toward this objective has been implemented and tested successfully in a number of practical domains. The research project comprises of three parts centering around rule- based connectionist networks. The first is to investigate an inductive learning method which combines neural learning and symbolic artificial intelligence techniques. The second is to investigate the knowledge-based neural network approach for revising existing knowledge and learning new knowledge. The third is to investigate an incremental rule-based connectionist model for machine learning.//
