Pattern Recognition Using Distributed Fuzzy Rules

In this paper, we propose the concept of distributed representations of fuzzy if-then rules and apply it to classification problems. The distributed representations are implemented by superimposing many fuzzy rules corresponding to different fuzzy partitions of an input space. First we compare the distributed representations with the conventional fuzzy rules which can be viewed as local representations. Next we propose an algorithm to construct fuzzy rules automatically from given data for two-group discriminant problems. The proposed method requires neither adjustment nor learning of membership functions. We also propose a fuzzy reasoning method using the distributed fuzzy rules. The proposed method are illustrated by numerical examples. Last we demonstrate the classification power of the distributed fuzzy rules using the classification problem of iris in Fisher.