Reproduce “Schrödinger Bridge Matching for Tree-Structured Costs and Entropic Wasserstein Barycentres” 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
Samuel Howard, Peter Potaptchik, George Deligiannidis
TreeDSBM extends Iterative Markovian Fitting (IMF) to solve Schrödinger bridges on tree-structured costs, including Wasserstein barycentres. It alternates Brownian bridge sampling with learned bridge-matching updates, avoiding repeated full transport-map solves. On a two-dimensional moon, spiral, and circle task, six IMF iterations achieved Sinkhorn divergence 1.14 ± 0.07 versus 2.35 for TreeDSB after 50 iterations; this is one approximate-reference comparison.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.