Bayesian Identification of Non-stationary Ar Model with Unknown Forgetting Factor

In this paper, we study Bayesian identification of the nonstationary parameters of the AR process. It is traditionally achieved via forgetting. Numerically efficient solution is available if the forgetting factor is known a priori. In this paper, we propose a joint Bayesian identification of the AR parameters and the unknown forgetting factor. The resulting intractable posterior distribution is approximated using Variational-Bayes method. Illustration of the method on simple simulated data is presented.