Reproduce “Generalizing Metropolis-Hastings for DAG Sampling” 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

Daniele Nikzad, Mikko Koivisto

The authors introduce duet sampling for Bayesian network structure learning, using paired proposal rules to generate a directed acyclic graph candidate and evaluate the reverse move while preserving the posterior. Their unconstrained sequential resampling move updates several parent sets; reverse-order pairing can favor changes in ancestor relationships. On Hailfinder, USR-rev achieved WEPD 0.0074 versus 1.00 for Gibby-rev, indicating run agreement, not exact posterior accuracy.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.