Reproduce “Closing the Approximation Gap in Simulation-free Latent SDEs” 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
Henry D. Smith, Brian L. Trippe, Scott W. Linderman
Helmholtz-SDE adds a divergence-free correction to simulation-free inference for latent stochastic differential equations, allowing the evidence lower bound to select among path laws with the same one-time marginals. In predator–prey data, its learned period and predator lag were 6.92 and 1.66 days, near empirical estimates; exact recovery is proved only for linear-Gaussian conjugate settings.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.