Reproduce “Anchored Stein Variational Gradient Descent Algorithm for Non-smooth 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
Thanh Dang, Matthew X. Burns, Jiaming Liang
Anchored SVGD samples targets with locally Lipschitz potentials without computing their gradients, using a smoother reference potential, its gradient, and weights based on target-potential values. The authors derive its KL steepest-descent direction and prove a conditional bound on expected, iteration-averaged squared anchored KSD. The theorem requires stated regularity and initialization conditions; it does not guarantee target-metric convergence.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.