Reproduce “Distributed Bayesian Learning with Stochastic Natural-gradient Expectation Propagation and the Posterior Server” 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
Leonard Hasenclever, Stefan Webb, Thibaut Lienart, Sebastian Vollmer, Balaji Lakshminarayanan, Charles Blundell, Yee Whye Teh
SNEP enables Bayesian learning when data are split across workers, using a posterior server to coordinate an approximation to the shared parameter distribution. Workers sample from local distributions and update likelihood factors through stochastic natural-gradient steps based on sampled moments. In simulated logistic regression, SNEP was more stable, faster, and more accurate than SMS, especially with few samples; this comparison does not establish broader performance or privacy guarantees.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.