Reproduce “Provable Benefit of Annealed Langevin Monte Carlo for Non-log-concave 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

Wei Guo, Molei Tao, Yongxin Chen

The authors introduce annealed Langevin Monte Carlo (ALMC), gradually changing an easy distribution into a target while taking Langevin steps. The bound depends on path action and requires no log-concavity or isoperimetry. Under assumptions, oracle calls scale as Õ(dβ²A²ε⁻⁶) for KL error at most ε². In a six-peak test, iterations grew roughly as a power of radius; this is not a guarantee for all targets.

Summary by alphaXiv.

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