Reproduce “Analyzing and Improving Optimal-Transport-based Adversarial Networks” 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
Jaemoo Choi, Jaewoong Choi, Myungjoo Kang
The authors introduce UOTM-SD, which schedules the distribution-matching weight during training to address UOTM’s sensitivity to the transport-cost weight τ. The method begins with weaker distribution-matching pressure and increases it over time. On CIFAR-10 at 32×32, its linear schedule reports a Fréchet Inception Distance (FID) of 2.51 versus 2.97 for UOTM; the theorem covers transport plans, not learned maps.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.