Extracting heuristically acceptable information from fuzzy/neural architectures via heuristic constraint enforcement. I. Foundation

Knowledge extraction from systems where the existing knowledge is limited is a difficult task. Using fuzzy/neural architectures to extract heuristic information from systems has received increasing attention. In most cases, using output error measures to validate extracted knowledge is not sufficient; extracted knowledge may not make heuristic sense even if the output error may meet the specified criterion. Using the principles of set theoretic estimation, the paper proposes a method for enforcing heuristic constraints on the membership functions of fuzzy/neural architectures. The proposed method ensures that the final membership functions conform to a priori heuristic knowledge. Although the method is described on a specific fuzzy/neural architecture, it is applicable to other realizations of fuzzy inference systems including adaptive or static implementations. The organized yet flexible characteristic of the heuristic constraint enforcement method enables its application to a wide range of problems.