Reproduce “Preferential Subsampling for Stochastic Gradient Langevin Dynamics” 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
Srshti Putcha, Christopher Nemeth, Paul Fearnhead
Authors develop preferential subsampling for stochastic-gradient Langevin dynamics, reweighting sampled observations to keep gradient estimates unbiased. For control-variate sampling, they also adapt minibatch size to the chain’s distance from a reference mode. On balanced synthetic logistic regression, adaptive sampling processes less data over 10,000 iterations without a measured reduction in accuracy; results do not establish this pattern universally.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.