Monge–Kantorovich Optimal Transport

• The Monge–Kantorovich optimal transport problem is defined as minimizing the cost of transporting mass between measures using both Monge’s mapping and Kantorovich’s relaxation techniques.
• It employs duality and twist conditions, such as m-twist and generalized-twist, to decompose optimal plans into finite or countable graph structures that ensure solution uniqueness.
• Applications span geometry, partial differential equations, economics, machine learning, and quantum information, providing a versatile framework for optimal resource allocation.

The Monge–Kantorovich optimal transport problem is a foundational concept in mathematical analysis, probability, and optimization, addressing the task of transporting mass or distributions between spaces in a cost-minimizing way. This theory underpins numerous advances in geometry, partial differential equations, economics, machine learning, quantum information, and signal processing.

1. Definitions and Primal–Dual Structure

Let $(X,d_X)$ and $(Y,d_Y)$ be Polish spaces and $\mu\in\mathcal{P}(X)$, $\nu\in\mathcal{P}(Y)$ Borel probability measures. Consider a continuous cost $c:X\times Y\to\mathbb{R}$.

Monge’s problem seeks a measurable map $T:X\to Y$ (the transport map) such that $T_\#\mu=\nu$ and

$$\inf_{T_\#\mu=\nu}\int_X c\big(x,T(x)\big)\,d\mu(x).$$

Kantorovich’s relaxation allows for transport plans $\pi\in\mathcal{P}(X\times Y)$ with marginals $\mu,\nu$: $(Y,d_Y)$0 Here, $(Y,d_Y)$1 is the set of Borel probability measures on $(Y,d_Y)$2 with specified marginals.

Dual formulation: Introduce potentials $(Y,d_Y)$3 with

$(Y,d_Y)$4

Kantorovich duality yields: $(Y,d_Y)$5 For bounded below, lower semicontinuous costs, the supremum is attained for $(Y,d_Y)$6-concave functions $(Y,d_Y)$7 (Moameni, 2014, Guillen et al., 2010).

2. Structural Results: Twist Conditions and Graph Decomposition

m-Twist and Generalized-Twist

Assume $(Y,d_Y)$8 is $(Y,d_Y)$9 in $\mu\in\mathcal{P}(X)$0. For each $\mu\in\mathcal{P}(X)$1, define

$\mu\in\mathcal{P}(X)$2

$\mu\in\mathcal{P}(X)$3

Theorem 1.3 (finite-graph decomposition under m-twist):

Let $\mu\in\mathcal{P}(X)$4 be a separable Riemannian manifold, $\mu\in\mathcal{P}(X)$5 non-atomic, $\mu\in\mathcal{P}(X)$6 bounded continuous and m-twist, every $\mu\in\mathcal{P}(X)$7-concave function differentiable $\mu\in\mathcal{P}(X)$8-a.e. Then every optimal plan admits a decomposition: $\mu\in\mathcal{P}(X)$9 for measurable maps $\nu\in\mathcal{P}(Y)$0, nonnegative weights $\nu\in\mathcal{P}(Y)$1 (Moameni, 2014).

Theorem 1.4 (countable-graph under generalized-twist):

The same holds with a countable decomposition for generalized-twist.

This result extends Brenier–McCann’s uniqueness theorem for strictly twisted costs.

3. Uniqueness Criteria and Support

When the support of an optimal plan is a finite union of graphs, refined uniqueness results are available.

Theorem 4.1 (uniqueness for finitely many graphs):

Given measurable maps $\nu\in\mathcal{P}(Y)$2 with injectivity and disjointness of ranges (except possibly for $\nu\in\mathcal{P}(Y)$3), and a separation function $\nu\in\mathcal{P}(Y)$4 such that

$\nu\in\mathcal{P}(Y)$5

there is at most one $\nu\in\mathcal{P}(Y)$6 supported on $\nu\in\mathcal{P}(Y)$7 (Moameni, 2014).

Consequently, under strict twist (m=1), or when the convex decomposition collapses to one map, the solution is a unique Monge map; for m-twist, at most m maps appear in the decomposition.

4. Measure-Theoretic and Proof Perspectives

A central tool is the measure-theoretic selection principle and Choquet representation on the map $\nu\in\mathcal{P}(Y)$8. The convex set of preimages of $\nu\in\mathcal{P}(Y)$9 admits an extreme-point decomposition (von Weizsäcker–Winkler theory):

• Each extreme section corresponds to a measurable selection, and the m-twist condition limits the number of these to m (Moameni, 2014).
• By Kantorovich duality, optimal plans relate to $c:X\times Y\to\mathbb{R}$0-concave potentials, whose gradients relate to the support: $c:X\times Y\to\mathbb{R}$1 $c:X\times Y\to\mathbb{R}$2-a.e.
• For the uniqueness theorem, the argument partitions the support into aperiodic unions and reduces uniqueness to that for doubly stochastic measures on such unions.

5. Illustrative Examples and Applications

• Circle cost: $c:X\times Y\to\mathbb{R}$3 on $c:X\times Y\to\mathbb{R}$4 is 2-twist; for $c:X\times Y\to\mathbb{R}$5, the optimizer is supported on two graphs.
• Smooth manifolds: A $c:X\times Y\to\mathbb{R}$6 cost $c:X\times Y\to\mathbb{R}$7 with everywhere non-singular mixed Hessian $c:X\times Y\to\mathbb{R}$8 is locally 1-twisted, which recovers classical results that optimal plans are supported on Lipschitz submanifolds.
• Classical case: Under 1-twist and absolute continuity of $c:X\times Y\to\mathbb{R}$9, the unique solution is a Monge map, matching the Brenier–McCann theorem (Moameni, 2014).

Applications

• Boundary Matching in Computer Vision: Stratified Monge–Kantorovich problems address cases where sources and/or targets live on lower-dimensional sets, such as the interior/boundary decomposition relevant to shape recognition (Ahmadpoor et al., 2024).
• Extensions to Multi-Marginal Settings: Recent theory generalizes the above characterization to multi-marginal optimal transport problems, replacing classical twist with twistedness on c-splitting sets, and yield analogous finite or countable graph representations of optimizers (Moameni, 2014).

6. Generalizations and Future Directions

The measure-theoretic and duality-based framework for the Monge–Kantorovich problem allows for significant extensions:

• Stratified OT: Existence and uniqueness even under singular marginals supported on stratified or multi-layered sets, combining absolutely continuous and singular parts (Ahmadpoor et al., 2024).
• Graph Decomposition: For non-degenerate costs (having locally finite pre-images of the gradient in $T:X\to Y$0), the generalized-twist condition applies and the support decomposes accordingly.
• Martingale and Vector-valued OT: The abstract duality approach supports further generalizations, such as vector measure versions or problems with martingale constraints (Gover, 23 Jan 2025).
• Numerical Methods: Discretizations and adaptive schemes—such as monotone wide-stencil finite-difference and entropic-regularized linear programming—allow practical computation of optimal maps and plans, with convergence guarantees to the continuum problem (Benamou et al., 2017, Solomon, 2018).
• Quantum and Matrix-valued OT: Quantum analogues define transport between density matrices with cost operators and semidefinite constraints; matrix-valued formulations allow spectral transport with rotation cost terms (Friedland et al., 2021, Ning et al., 2013).

7. References to Key Results

Topic	Main Theorems/Conditions	Reference
m-twist & decomposition	Theorems 1.3, 1.4: Graph decomposition, support structure	(Moameni, 2014)
Uniqueness	Theorem 4.1: Uniqueness for support on union of graphs	(Moameni, 2014)
Stratified, Lower-dim	Existence and decomposition for layers/mixed support	(Ahmadpoor et al., 2024)
Generalized-twist	Link to non-degenerate cost, support, existence	(Moameni, 2014)

This measure-theoretic characterization, leveraged with duality, provides a unified description of structure and uniqueness for optimal transport plans in a broad range of settings, encompassing and extending classical results. The support of optimal plans, governed by twist-type conditions, may comprise multiple graphs, but is always sharply controlled by the regularity—and degeneracy—properties of the cost.