Reproduce “Neural Backward Filtering Forward Guiding” 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
Gefan Yang, Frank van der Meulen, Stefan Sommer
NBFFG reconstructs states in branching processes from observations at the leaves, using a tractable proxy to send descendant information backward and a learned correction to guide proposals forward. Its variational objective still uses the true model. On a synthetic tree, correction reduced average marginal KL from 1.861 ± 0.096 to 0.118 ± 0.004 over five runs; this does not validate butterfly ancestors.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.