BART

BART (Bayesian Additive Regression Trees) is a nonparametric model that inherits the expressiveness of decision tree ensembles and the uncertainty quantification of Bayesian methods. As is often the case, the posterior distribution of this model does not have a closed-form expression, so Markov Chain Monte Carlo (MCMC) methods are used to work with it.

We will discuss the structure of the model as well as the MCMC methods used. In the remaining time, we will also cover directions for further research. The material does not require specialized background knowledge, but familiarity with basic concepts of Markov chain theory up to stationary distributions will be helpful.