Code for Bayesian Additive Regression Spanning Trees, a tool for nonparametric regression on complex constrained domains.

Reference:

Luo, Z. T., Sang, H., & Mallick, B. (2021) BAST: Bayesian Additive Regression Spanning Trees for Complex Constrained Domain. Advances in Neural Information Processing Systems 34 (NeurIPS 2021)

BASTDemo.R: Demonstration code of BAST

AralSea.R: Preprocessing chlorophyll data in Aral Sea for demonstration purposes, following Niu et al. (2019)

BASTFun.R: Implementation and API documentations for main functions of BAST

ComplexDomainFun.R: Utility functions for complex constrained domains

FEMFun.R: Utility functions for triangular meshes, modified from R package fdaPDE

aral_data.RData: Preprocessed data from AralSea.R

The code depends on the following R packages: igraph, fields, FNN, mgcv, fdaPDE, gamair. Please make sure they are installed before running the demo code.

A more efficient version using Rcpp is under active development.