Reproduce “Austerity in MCMC Land: Cutting the Metropolis-Hastings Budget” 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

Anoop Korattikara, Yutian Chen, Max Welling

The authors introduce a Metropolis–Hastings method that decides proposal acceptance by examining data in batches and stopping when evidence suffices. This reduces computation per decision but introduces bias. In MNIST logistic regression distinguishing digits 7 and 9, a nonzero-bias setting had the lowest plotted finite-time risk for predictive-mean estimates on 2,037 test points; exact MH is expected to outperform it with infinite time.

Summary by alphaXiv.

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