The paper introduces BART (Bayesian Additive Regression Trees), a Bayesian "sum-of-trees" model where each tree is constrained by a regularization prior to be a weak learner. The model is fitted and inferred using an iterative Bayesian backfitting MCMC algorithm that generates samples from the posterior distribution. BART is a nonparametric Bayesian regression approach that uses dimensionally adaptive random basis elements. It is defined by a statistical model consisting of a prior and a likelihood, enabling full posterior inference, including point and interval estimates of the unknown regression function and marginal effects of potential predictors. BART can also be used for model-free variable selection by tracking the frequency of predictor inclusion in the sum-of-trees model iterations. The paper includes a bake-off against competing methods on 42 different datasets, a simulation experiment, and a drug discovery classification problem to illustrate BART's performance. The authors provide open-source software implementing BART as the **BayesTree** library in R.The paper introduces BART (Bayesian Additive Regression Trees), a Bayesian "sum-of-trees" model where each tree is constrained by a regularization prior to be a weak learner. The model is fitted and inferred using an iterative Bayesian backfitting MCMC algorithm that generates samples from the posterior distribution. BART is a nonparametric Bayesian regression approach that uses dimensionally adaptive random basis elements. It is defined by a statistical model consisting of a prior and a likelihood, enabling full posterior inference, including point and interval estimates of the unknown regression function and marginal effects of potential predictors. BART can also be used for model-free variable selection by tracking the frequency of predictor inclusion in the sum-of-trees model iterations. The paper includes a bake-off against competing methods on 42 different datasets, a simulation experiment, and a drug discovery classification problem to illustrate BART's performance. The authors provide open-source software implementing BART as the **BayesTree** library in R.