Reproduce “Accelerating Hamiltonian Monte Carlo via Chebyshev Integration Time” 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
Jun-Kun Wang, Andre Wibisono
The paper introduces a variable-duration Hamiltonian Monte Carlo schedule that uses shifted Chebyshev-polynomial roots to choose trajectory lengths for distributions whose directions have different curvature. For ideal HMC on Gaussian targets, it proves O(√κ log(1/ε)) iterations, versus O(κ log(1/ε)) for constant duration. Experiments report higher effective sample sizes per CPU second, but acceleration is unproven for general strongly log-concave targets.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.