Neural network system identification for improved noise rejection

From MaRDI portal





The paper deals with the noise reduction problem in linear system identification. The development relies on the use of a batch ARMarkov model, a generalization of an ARMA model, and implements a neural network model for solving the problem. It is shown that the neural network from a batch ARMarkov model is able to improve the noise rejection properties, and that asymptotically (with model parameter values growing to infinity) the scheme is capable of identifying the system with arbitrary accuracy. An array of neurons with linear activation functions is used as the basic building block of the neural network and the evolution rules of the corresponding weighting matrices are consequently derived by means of matrix algebra. Some illustrative examples, concerning identification problems with white and coloured noise, are included.











This page was built for publication: Neural network system identification for improved noise rejection

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4364036)