On-line Bayesian system identification

We consider an on-line system identification scenario in which new data become available at given times. In order to meet real-time estimation requirements, we propose a tailored Bayesian system identification procedure in which the hyper-parameters are estimated through one-step-updates of an algorithm optimizing the Marginal Likelihood. To this purpose both gradient methods and an EM algorithm are considered. We compare this “1-step” procedure with the standard one, in which the optimization method is run until convergence to a local minimum. The experiments confirm the effectiveness of this approach.