Reproduce “Path Integral Sampler: a stochastic control approach for 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
Qinsheng Zhang, Yongxin Chen
Path Integral Sampler (PIS) learns a target-specific stochastic process to sample from densities known up to a normalizing constant. It guides noisy particles from a simple starting distribution with a trained, time-varying drift, using a stochastic-control objective. On a 10-dimensional Funnel task, PIS-RW-Grad reports combined error 0.012 versus AFT’s 0.284; learned finite-step sampling is not guaranteed optimal.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.