Reproduce “A Unifying View of Distributionally Robust Optimization for Probability Smoothing via Maximum Entropy” 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
Kota Kurihara, Yuya Yamakawa, Yoichi Izunaga
OT-DRO smooths finite-sample probability estimates using an optimal-transport cost matrix, so category relationships guide probability assigned to unseen categories. Under stated conditions, the authors show that distributionally robust smoothing is maximum-entropy within its ambiguity set. In synthetic tests with 100 categories and 100 draws, OT-DRO lowers mean KL divergence versus 2-DRO in structured data but raises it in unstructured data.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.