Reproduce “Hyperbolic 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

Yan Bin Ng, Xianfeng Gu

Researchers at Stony Brook University developed a robust method for computing optimal transport maps in hyperbolic space by extending geometric variational principles and hyperbolic Legendre duality. This approach efficiently and accurately compares probability distributions on negatively curved manifolds, demonstrating computational times comparable to Euclidean optimal transport methods on standard hardware.

Key insights

  • The hyperbolic optimal transport problem can be rigorously recast as a generalized Minkowski problem in Minkowski spacetime through the generalization of convex geometry and Legendre duality to hyperbolic settings.
  • The Kantorovich functional associated with hyperbolic optimal transport is concave and twice-differentiable, which enables efficient and robust optimization using gradient-based methods like Newton's method.
  • Hyperbolic power diagrams provide a practical and theoretically sound geometric tool for constructing optimal transport cells and deriving their differential properties in hyperbolic space, crucial for numerical implementation.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.