Recursive structure estimation for nonlinear identification with modular networks
Visakan Kadirkamanathan, Simon G. Fabri · 2002
The paper presents a recursive nonlinear identification scheme with modular networks consisting of local linear models. New local linear models are added online as and when necessary. The algorithms is developed within a probabilistic framework and utilises the Kalman filter for estimation of model parameters. Simulated results demonstrate the operation of the algorithm.