The paper introduces a new way to measure distances between probability distributions by blending two existing mathematical frameworks. The first, called Wasserstein geometry, captures how much "effort" it takes to physically move mass from one distribution to another, like rearranging piles of sand. The second, called Aitchison geometry, treats distributions as compositional data, meaning it cares about the relative proportions of mass rather than its absolute location. The authors combine these into a single hybrid distance that simultaneously accounts for spatial transport and changes in the shape or composition of the distribution. They prove this hybrid distance satisfies all the formal requirements to be a true metric, and they characterize how strong its notion of closeness is compared to other standard topologies used in analysis.
Beyond just defining the distance, the authors build a substantial mathematical theory around it. A central tool in modern analysis is the idea of a "gradient flow," which describes how a system evolves over time to minimize some energy, much like water flowing downhill. The authors establish the technical foundations needed to run this theory in their new geometric setting, including precise notions of how curves behave, how fast they move, and how steep the energy landscape is. They also study the notion of a "barycenter," which is a kind of weighted average of distributions in the hybrid geometry, and connect it to analogous concepts in the two parent frameworks.
The payoff is a fresh perspective on several important partial differential equations that describe physical and biological phenomena. Equations governing logistic diffusion, phase separation with compositional constraints, and chemotaxis with logarithmic growth can all be reinterpreted as gradient flows within this hybrid geometry. This reinterpretation is valuable because gradient flow structures often reveal hidden mathematical properties, suggest natural numerical schemes, and provide stability and uniqueness results. The authors also compare their construction to an existing competitor called the Wasserstein-Fisher-Rao metric, clarifying where the new framework offers a genuinely different viewpoint.