Department Seminar Series: Nicholas Polson, Recursive Bayesian Computation
Abstract: In this paper we develop a framework for recursive Bayesian computation. By exploiting an auxiliary latent variable structure we provide sequential parameter learning for a wide class of models. We illustrate our methodology with applications to high dimensional sparse regression, dynamic logistic classification, mixture Kalman filters and nonlinear and non-Gaussian state space models. The methods developed here are available in the package ParticleBayes.R.