Recurrent neural networks for recursive identification of nonlinear dynamic process
Jia-Wen Dong, Jixin Qian, Youxian Sun · 2002
In this paper, modified Elman-type recurrent neural networks (1990) were developed to identify the dynamic nonlinear systems with generalised backpropagation recursive algorithm. Analysis shows that introduction of adjustable self-connections of context units provides network ability to model high order input-output mapping, unbiased estimates can be achieved without the need to fit additive noise model. An industrial application example shows its efficiency.>