Bayesian methods for autoregressive models

We describe a variational Bayesian (VB) learning algorithm for parameter estimation and model order selection in antoregressive (AR) models. With uninformative priors on the precisions of the coefficient and noise distributions the VB framework is shown to be identical to the Bayesian evidence framework. The VB model order selection criterion is compared with the minimum description length (MDL) criterion on synthetic data and on EEG.

[1]  Athanasios Papoulis,et al.  Probability, Random Variables and Stochastic Processes , 1965 .

[2]  T. Hassard,et al.  Applied Linear Regression , 2005 .

[3]  Thomas M. Cover,et al.  Elements of Information Theory , 2005 .

[4]  Petar M. Djuric,et al.  Order selection of autoregressive models , 1992, IEEE Trans. Signal Process..

[5]  Piet M. T. Broersen,et al.  Autoregressive model order selection by a finite sample estimator for the Kullback-Leibler discrepancy , 1998, IEEE Trans. Signal Process..

[6]  Hagai Attias,et al.  A Variational Bayesian Framework for Graphical Models , 1999 .

[7]  Charles M. Bishop.  Variational principal components , 1999 .

[8]  S. Roberts,et al.  Variational Bayes for non-Gaussian autoregressive models , 2000, Neural Networks for Signal Processing X. Proceedings of the 2000 IEEE Signal Processing Society Workshop (Cat. No.00TH8501).