Coupling Matrix Manifolds and Their Applications in Optimal Transport
This work provides a novel geometric perspective for solving optimal transport problems, which is incremental but offers improved performance in specific scenarios.
The paper introduces the coupling matrix manifold (CMM) as a new geometric framework for optimal transport problems, enabling the development of Riemannian optimization algorithms that perform comparably to classic methods like Sinkhorn and outperform other state-of-the-art approaches, particularly in non-entropy optimal transport cases.
Optimal transport (OT) is a powerful tool for measuring the distance between two defined probability distributions. In this paper, we develop a new manifold named the coupling matrix manifold (CMM), where each point on CMM can be regarded as the transportation plan of the OT problem. We firstly explore the Riemannian geometry of CMM with the metric expressed by the Fisher information. These geometrical features of CMM have paved the way for developing numerical Riemannian optimization algorithms such as Riemannian gradient descent and Riemannian trust-region algorithms, forming a uniform optimization method for all types of OT problems. The proposed method is then applied to solve several OT problems studied by previous literature. The results of the numerical experiments illustrate that the optimization algorithms that are based on the method proposed in this paper are comparable to the classic ones, for example, the Sinkhorn algorithm, while outperforming other state-of-the-art algorithms without considering the geometry information, especially in the case of non-entropy optimal transport.