Applying geometric compatibility modification in FuzzyCLIPS
Valerie V. Cross, A. Rajagopal · 2002
A new form of fuzzy reasoning, geometric compatibility modification (GCM) inference, is currently under investigation through its implementation within FuzzyCLIPS, a fuzzy systems development tool from the National Research Council of Canada. Using the dissemblance compatibility measure, GCM inference is able to determine a conclusion or an action even when the fuzzy input does not intersect the fuzzy antecedent of any rule. This situation can occur in sparse rule bases. The initial goal of GCM implementation within FuzzyCLIPS is to compare its performance to the standard fuzzy inference technique with compact rule bases. The next goal is to capitalize on GCM's ability to infer a consequent even in sparse rule bases where conventional fuzzy reasoning methods are unable to derive one. The results and knowledge gained from the FuzzyCLIPS GCM implementation, and testing with a simple control problem are presented. The future plans for continued investigation of GCM inference are discussed.