Lt. Koczy et K. Hirota, APPROXIMATE REASONING BY LINEAR RULE INTERPOLATION AND GENERAL APPROXIMATION, International journal of approximate reasoning, 9(3), 1993, pp. 197-225
The problem of sparse fuzzy rule bases is introduced Because of the hi
gh computational complexity of the original compositional rule of infe
rence (CRI) method, it is strongly suggested that the number of rules
in a final fuzzy knowledge base is drastically reduced. Various method
s of analogical reasoning available in the literature are reviewed. Th
e mapping style interpretation of fuzzy rules leads to the idea of app
roximating the fuzzy mapping by using classical approximation techniqu
es. Graduality, measurability, and distance in the fuzzy sense are int
roduced. These notions enable the introduction of the concept of simil
arity between two fuzzy terms, by their closeness derived from their d
istance. The fundamental equation of linear rule interpolation is give
n, its solution gives the final formulas used for interpolating pairs
of rules by their alpha-cuts, using the resolution principle. The meth
od is extended to multiple dimensional variable spaces, by the normali
zation of all dimensions. Finally, some further methods are shown that
generalize the previous idea, where. various approximation techniques
are used for the alpha-cuts and so, various approximations of the fuz
zy mapping R: X --> Y.