Semi-discrete unbalanced optimal transport and quantization

David P. Bourne,Bernhard Schmitzer,Benedikt Wirth

2024-07-16

Abstract:In this paper we study the class of optimal entropy-transport problems introduced by Liero, Mielke and Savaré in Inventiones Mathematicae 211 in 2018. This class of unbalanced transport metrics allows for transport between measures of different total mass, unlike classical optimal transport where both measures must have the same total mass. In particular, we develop the theory for the important subclass of semi-discrete unbalanced transport problems, where one of the measures is diffuse (absolutely continuous with respect to the Lebesgue measure) and the other is discrete (a sum of Dirac masses). We characterize the optimal solutions and show they can be written in terms of generalized Laguerre diagrams. We use this to develop an efficient method for solving the semi-discrete unbalanced transport problem numerically. As an application we study the unbalanced quantization problem, where one looks for the best approximation of a diffuse measure by a discrete measure with respect to an unbalanced transport metric. We prove a type of crystallization result in two dimensions -- optimality of a locally triangular lattice with spatially varying density -- and compute the asymptotic quantization error as the number of Dirac masses tends to infinity.

Optimization and Control

What problem does this paper attempt to address?

### Problems Addressed by the Paper

This paper primarily investigates the semi-discrete unbalanced optimal transport problem and its applications in quantization problems. Specifically:

1. **Semi-Discrete Unbalanced Transport Theory**:
- The paper develops the theory of semi-discrete unbalanced transport problems, where one measure is diffuse (absolutely continuous) and the other measure is discrete, and the total masses of the two measures can be different (unbalanced).
- It studies the properties of the optimal solutions and proves that these solutions can be represented by generalized Laguerre diagrams.
- It proposes an efficient numerical method to solve such semi-discrete unbalanced transport problems.

2. **Unbalanced Quantization Problem**:
- The above theory is applied to the unbalanced quantization problem, which involves approximating a diffuse measure with a discrete measure such that the error under the unbalanced transport metric is minimized.
- In 2-dimensional space, a lattice result is proven, showing that a local triangular lattice is optimal in large-scale cases.
- The asymptotic quantization error is calculated as the number of discrete points tends to infinity.

### Main Contributions

1. **Theoretical Extensions**:
- The semi-discrete transport theory is extended to the unbalanced case, providing geometric partition formulas (Theorem 3.1) and optimality conditions (Theorem 3.2).
- Numerical algorithms are developed to solve the semi-discrete unbalanced transport problem, showcasing new phenomena in unbalanced transport.

2. **Applications to Quantization Problems**:
- The optimal transport quantization theory is extended to unbalanced transport, resulting in an equivalent Voronoi partition problem (Theorem 4.1).
- Unbalanced quantization is demonstrated numerically, and standard algorithms (including the Lloyd algorithm) are extended to the unbalanced case.

3. **Asymptotic Analysis**:
- In 2-dimensional space, the optimal asymptotic quantization cost and asymptotic point density under unbalanced transport are derived (Theorem 4.14).
- The classical Zador theorem is included as a special case, and the non-local dependence between point density and global mass distribution in the unbalanced setting is demonstrated.

Through these studies, the paper not only advances the development of semi-discrete unbalanced transport theory but also provides new theoretical foundations and numerical methods for unbalanced quantization problems.