Incremental locally linear fuzzy modeling

In this paper we propose an approach to incrementally build fuzzy models based on two principles. Principle I Put only rules where they are needed. Rules are placed where they contribute most to model the underlying function. Given the fact that a fuzzy model basically behaves like a piecewise linear interpolator, rules should be put at points where there is a non-linearity. Principle II Use tetrahedron shaped fuzzy antecedents to cope with the curse of dimensionality.