Reproduce “Distributionally Robust Schrödinger Bridge” 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
Jinhwan Sul, Panagiotis Theodoropoulos, Vincent Pacelli, Jaemoo Choi, Evangelos Theodorou
The authors introduce DRSB, which trains one controller to transport samples from a nominal starting distribution to a fixed target while accounting for plausible shifts in the starting distribution. It minimizes worst-case control energy plus a terminal-mismatch penalty. On GMM-to-GMM transport, DRSB-S has lower mean SWD than fixed-noise augmentation at unseen noise levels δ=3 and δ=4, but worse nominal-input SWD.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.