Reproduce “Robust Approximate Sampling via Stochastic Gradient Barker 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
Lorenzo Mauri, Giacomo Zanella
The authors extend Barker proposal to minibatch gradients, creating SGBD, an unadjusted sampler in which gradients steer move directions rather than set their magnitudes. They correct nonlinear sign-choice probability under Gaussian gradient noise, with a per-coordinate error bound of 0.019 under stated conditions. Experiments report greater robustness than SGLD to step-size choices and irregular geometry, not guaranteed posterior accuracy.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.