Reproduce “DynamicPPL: Stan-like Speed for Dynamic Probabilistic Models” 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
Mohamed Tarek, Kai Xu, Martin Trapp, Hong Ge, Zoubin Ghahramani
DynamicPPL uses trace-based type inference to make dynamic probabilistic models amenable to efficient execution. It first records variables in a flexible trace, then uses inferred types in a concretely typed trace that Julia can specialize. In eight benchmarks, Turing with AdvancedHMC was faster than Stan on three and slower on five; logistic regression took 4.225 versus 60.932 seconds, without isolating DynamicPPL's contribution.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.