Parameter estimation of state space models by recurrent neural networks

Four variants of recurrent neural networks (RNNs) are studied. The similarities and contradistinction of these formulations are brought out from the view point of their applicability to parameter estimation in dynamic systems. The trajectory matching algorithms are also given. A recursive information processing scheme within the structure of a Hopfield neural network for parameter estimation is presented. Numerical simulation results for nonrecursive and recursive schemes are given