UvA

This paper presents an approach to similarity based approximate reasoning that elucidates the connection between similarity and existing approaches to inference in approximate reasoning methodology. A set of axioms is proposed to get a reasonable measure of similarity between two fuzzy sets. The similarity between the fact(s) and the antecedent of a rule is used to modify the relation between the antecedent and the consequent of the rule. An inference is drawn using the well-known projection operation on the domain of the consequent. Zadeh's compositional rule of inference and existing similarity based reasoning techniques are considered for a new similarity based approximate reasoning technique. The proposed mechanism is used to develop a modified fuzzy control system. A new defuzzification scheme is proposed. Simulation results are presented for the well-known inverted pendulum problem.