ABSTRACT
Fuzzy systems provide in general a non-linear mapping between the input and the output space. The parameters allowing the adjustment of this input/output mapping are related to the fuzzy partition of the input domain (including the shape of antecedent membership functions and their overlapping degrees), the partition of the output domain (the representation of the consequents), the antecedent connectives, the implication function, and the ways of implement ing rule aggregation and defuzzification (if needed). What is very remarkable is that any mapping between the input and the output space can be approximat ed to any degree of accuracy by properly choosing the above parameters. This means that fuzzy systems are universal approximators.
