Reproduce “Theoretical and Practical Advances in Hierarchical Bayesian Deep Learning Methodology” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Neil K. Chada, Simon Cotter, Kody J. H. Law, Xinzhu Liang
MLPMMH extends multilevel inference to joint estimation of weights and prior-scale hyperparameters in static Bayesian neural networks. Coupled sequential Monte Carlo estimates posterior differences across nested network widths, while a pseudo-marginal Metropolis–Hastings chain updates hyperparameters. In synthetic Bayesian regression, mean squared error decreased with computational cost at a fitted log–log rate of −0.9; this is not a measured comparison to ordinary PMMH.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.