Reproduce “Efficient Gradient-Based Inference through Transformations between Bayes Nets and Neural Nets” 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
Diederik P. Kingma, Max Welling
The authors introduce a differentiable non-centered parameterization that expresses latent states as functions of auxiliary noise and their parents. It preserves the Bayesian model and enables gradients through backpropagation. In a nonlinear dynamic Bayesian network at log σ_z = −5, robust HMC reached ESS 640, versus 2 for CP and 305 for DNCP, from 4,000 retained samples; this result is specific to that setting.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.