The structure of rules for general-purpose fuzzy reasoning

摘要

The structure desirable for rules in rule-based systems for fuzzy reasoning, as distinct from fuzzy control, is elaborated. We point out that the formal structure of rules for fuzzy control, while adequate and economical for most fuzzy control applications, is both too complex and too restrictive for optimal use in fuzzy reasoning systems; in particular, the difficulty faced in handling fuzzy sets of categories in attempting to answer questions of the "what is it?" type when using fuzzy control rules is noted. Fuzzy reasoning rules should be able to employ discrete fuzzy sets of linguistic values, with members such as "Slow" and "Large", and also discrete fuzzy sets of categories, with members such as "Jeep" and "Quartz". The importance in multistep logic of considering prior data values is mentioned, together with a way of handling prior information based on monotonic fuzzy reasoning.