(272a) Hybrid Neural-Network Kalman-Bucy Filter Approach for Adaptive Tracking of Complex Signals

A generic algorithm for signal tracking, based on an approximated continuous black-box Artificial Neural Network model as reference, is presented. The model is coupled with a Kalman-Bucy Filter as an adaptive corrector to improve its predictive capabilities. The hybrid models constructed this way are able to track the trajectories of complex systems responses even in the presence of noise. The approach is illustrated using a catalytic surface model with white noise as experimental system and experimental rumen dynamical data.