Non-Linear State estimation using pre-trained neural networks
Enis Bayramoglu, Nils Andersen, Ole Ravn, Niels Kjølstad Poulsen · 2010
This article presents a method to track non-Gaussian parametric probability density functions under nonlinear transformations and posterior calculations. The optimal set of parameters for the transformed distribution is a function of the parameters for the prior distribution and any other variables effecting the transformation. This function is approximated by a neural network using offline training. The training is based on monte carlo sampling. A way to obtain parametric distributions of flexible shape to be used easily with these networks is also presented. The method can also be used to improve other parametric methods around regions with strong non-linearities by including them inside the network.