ABSTRACT
The Bayesian Additive Regression Trees is a Bayesian non-parametric model that uses a sum of decision trees to obtain a flexible model, BARTs are a probabilistic version of Random Forest. The main Bayesian idea used by BARTs is that as decisions trees can easily overfit we add a regularizing prior to make each tree behave as a weak learner. By doing so we hope to get enough flexibility without too much complexity.
