English

Generative Regression with IQ-BART

Methodology 2025-07-08 v1 Machine Learning

Abstract

Implicit Quantile BART (IQ-BART) posits a non-parametric Bayesian model on the conditional quantile function, acting as a model over a conditional model for YY given XX. One of the key ingredients is augmenting the observed data {(Yi,Xi)}i=1n\{(Y_i,X_i)\}_{i=1}^n with uniformly sampled values τi\tau_i for 1in1\leq i\leq n which serve as training data for quantile function estimation. Using the fact that the location parameter μ\mu in a τ\tau-tilted asymmetric Laplace distribution corresponds to the τth\tau^{th} quantile, we build a check-loss likelihood targeting μ\mu as the parameter of interest. We equip the check-loss likelihood parametrized by μ=f(X,τ)\mu=f(X,\tau) with a BART prior on f()f(\cdot), allowing the conditional quantile function to vary both in XX and τ\tau. The posterior distribution over μ(τ,X)\mu(\tau,X) can be then distilled for estimation of the {\em entire quantile function} as well as for assessing uncertainty through the variation of posterior draws. Simulation-based predictive inference is immediately available through inverse transform sampling using the learned quantile function. The sum-of-trees structure over the conditional quantile function enables flexible distribution-free regression with theoretical guarantees. As a byproduct, we investigate posterior mean quantile estimator as an alternative to the routine sample (posterior mode) quantile estimator. We demonstrate the power of IQ-BART on time series forecasting datasets where IQ-BART can capture multimodality in predictive distributions that might be otherwise missed using traditional parametric approaches.

Keywords

Cite

@article{arxiv.2507.04168,
  title  = {Generative Regression with IQ-BART},
  author = {Sean O'Hagan and Veronika Ročková},
  journal= {arXiv preprint arXiv:2507.04168},
  year   = {2025}
}

Comments

48 pages, 7 figures

R2 v1 2026-07-01T03:47:55.949Z