2015 年 51 巻 3 号 p. 197-205
This paper is concerned with construction of non-parametric PWA (Piecewise Affine) models. Though their effectiveness is demonstrated in the literature, the existing identification method suffers from heavy computation load.This load is mainly due to the large data set which is required for representing the model. This paper proposes a method to construct non-parametric PWA models represented with less data points based on estimated Hessian.