Inference, inquiry and explanation in expert systems by means of fuzzy neural networks

Current research interests in hybrid architectures for intelligent systems focus on integrating the computational paradigms of expert systems and neural networks in a manner that exploits the strengths of both systems and expands the applications to which either system could be applied individually. Such systems, which are called connectionist expert systems, are discussed. The authors show how basic functions of expert systems, such as inference, inquiry and explanation, can be implemented by means of fuzzy neural networks. They introduce some measures of uncertainty to be used by the connectionist inference machine of a classification expert system when performing these functions. The emphasis is on the uncertainty processing by the different elements of the network