Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces BAND, a sparse Bayesian network approach for multivariate distribution estimation that addresses the curse of dimensionality by employing sparsity-aware conditional mean methods. The method achieves polynomial total variation convergence rates, significantly outperforming classical multivariate histogram density estimators, especially as the feature dimension increases with sample size. Empirical results demonstrate that BAND competes effectively with leading benchmarks in data sampling and confidence region forecasting tasks.
BAND achieves polynomial convergence rates in high-dimensional distribution estimation, breaking the curse of dimensionality that plagues traditional methods.
Minimax-optimal rates for multivariate distribution estimation are known to suffer from the curse of dimensionality. We propose a sparse Bayesian network approach in which each conditional probability is estimated using sparsity-aware conditional mean methods. The resulting estimator, \textit{BAyesian Network Distribution regression} (BAND), handles mixed data types in high-dimensional time series and achieves polynomial total variation convergence rates while allowing the feature dimension to grow polynomially with the sample size. These rates are substantially faster than the classical optimal rates for multivariate histogram density estimators that lack sparsity. Empirical evaluations show that BAND performs competitively for data sampling and confidence region forecasting against a range of state-of-the-art benchmarks.