Second generation fuzzy expert system

Abstract We describe our approach to designing a fuzzy expert systems shell which is decision theoretic in nature. We wish to determine the interval operating characteristics in order to maximize the information in relevant working memory. Our solution is a shell which generalizes, and contains, our present shell FLOPS. Major additions are: 1. (1) upper and lower confidence values; 2. (2) allowing rules to fire on either or both confidence values; and 3. (3) a memory update algorithm designed to maximize information.