Wrabel et al. approaches training a MLP for reducing the estimation error of the Kalman filter. To do so, the MLP is trained to