Fuzzy modeling using genetic algorithms with fuzzy entropy as conciseness measure (Q1602476)

A fuzzy modeling involving genetic algorithms combined with a new conciseness measure is described. This measure is nothing but the entropy \(-p\ln p-(1-p)\ln(1-p)\) first introduced in the literature by the Nobel winner Fermi and renamed later in the fuzzy literature. This entropy is used to evaluate the shape of a membership function. Nevertheless, since it could not distinguish between similarly shaped membership functions, one proposes a new measure for the deviation of a membership function from symmetry. Numerical results are displayed which show that the conciseness measure is in conflict with the accuracy of fuzzy models, and which the author has previously applied to fuzzy modeling using genetic algorithms.