Reproduce “Optimal Transport Meets Reinforcement Learning: A Survey” 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
Yujie Zhu, Charles A. Hepburn, Matthew Thorpe, Giovanni Montana
The authors survey how optimal transport (OT) compares distributions in reinforcement learning (RL), organizing methods by OT’s role, the distributions compared, formulation, and treatment of time. OT measures the cost of moving probability mass under a ground cost, which can help when distributions overlap weakly. Results are mixed: OTR+IQL approaches oracle-reward aggregate performance on D4RL locomotion, but not every task.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.