A way to reduce model complexity for Non Linear System Approximators
In this paper we exploit the approximation capabilities of non
linear function approximator based on fuzzy system structure, to devise an identification procedure for Single-Input Single-Output systems, which minimizes the squared error between the model and the target.
The adoption of a model featuring an increased locality allows a substantial reduction in the complexity of the identification phase in which samples are taken into account. Then, a data-independent mapping is devised to translate modified Non-Linear models into conventional ones.