Reproduce “A New Approach to Probabilistic Programming Inference” 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
Frank Wood, Jan Willem van de Meent, Vikash Mansinghka
Particle Markov chain Monte Carlo (PMCMC) uses Particle Gibbs to infer traces by repeatedly running particles, weighting them by observations, and retaining a trace between rounds. Unlike random database (RDB) Metropolis–Hastings, its proposals can change many choices. On tested HMM and DP-mixture tasks, it converged faster, though ordering can reverse the HMM comparison; this is not a general speedup guarantee.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.