Reproduce “Learning to Match via Inverse Optimal Transport” 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
Ruilin Li, Xiaojing Ye, Haomin Zhou, Hongyuan Zha
The Robust Inverse Optimal Transport (RIOT) framework infers adaptive, nonlinear matching cost functions from noisy empirical data by modeling observed matchings as outcomes of a regularized optimal transport problem. RIOT's novel marginal relaxation, utilizing regularized Wasserstein distance, provides enhanced robustness and superior predictive accuracy compared to existing methods in diverse real-world matching scenarios.
Key insights
- Marginal relaxation via regularized Wasserstein distance significantly improves the robustness of inverse optimal transport models when dealing with noisy or incomplete empirical matching data.
- Modeling matching costs with nonlinear kernel functions, rather than linear forms, allows for a more flexible and accurate representation of complex real-world preferences and interactions.
- By learning the underlying cost function, the framework provides interpretable insights into the factors driving observed empirical matchings, enabling prescriptive analytics beyond mere prediction.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.