Monotone Mamdani-type fuzzy systems with ellipsoidal antecedents

The paper deals with monotonicity conditions for Mamdani-type fuzzy systems where the antecedents of the rules are described by ellipsoidal regions. Such systems occur e. g. when clustering algorithms are used for partitioning the input data. The proposed algorithm is compared with non-monotonic fuzzy logic based multivariate regressor on few examples. The comparison shows that using the information about monotonicity may improve the regression.