Loss-based prior for CART and BART models

F. Serafini, F. Leisen, C. Villa +1

We present a novel prior for tree topology within Bayesian Additive Regression Trees (BART) models. This approach quantifies the hypothetical loss in information and the loss due t…