The Setting
The stability question compares quadratic optimal maps from a fixed uniform source when their target measures vary within a compact set. Let be Euclidean space, with a nonnegative integer, and let be compact and convex with nonempty interior. The uniform measure on the source body is the normalized restriction of Lebesgue measure in (1), where is the Lebesgue volume of : (1) A measure on is supported in a set when .✦
A coupling of measures on is a probability measure on with first marginal and second marginal ; write for the set of these couplings.✦ The quadratic transport cost of a measure on is the nonnegative extended real integral in (2): (2)✦ The quadratic Wasserstein distance between measures on is the square root of the real value of the infimum of over .✦
The graph transport plan of a map from is the pushforward in (3): (3)✦ A quadratic optimal transport map from to is a measurable map with whose graph plan minimizes over .✦ A unique quadratic optimal transport map from to is a quadratic optimal map such that every with equals .✦
The distance between maps is the square root of the integral in (4): (4)✦
A subgradient selection of a function is a map satisfying (5): (5)✦
The Claim
Theorem 1. For every , there is a constant with the following property. Let be continuous, and let be measurable subgradient selections of , respectively, with for every . If and some satisfies , then the normalized interpolation estimate is (6)✦
Let be nonnegative real numbers. If for every , then the slice stability estimate is (7)✦
Suppose and is compact and nonempty. There is a constant , depending only on , such that every pair of probability measures supported in admits unique quadratic optimal transport maps from satisfying (8)✦
In the paper, the result reads: for the global source and target setting, with bounding the norms on , respectively, the perimeter of , and , every pair of finitely supported probabilities on satisfies (9): (9)
What Is Checked
The assertions (6) and (7) concern continuous potentials with globally bounded measurable subgradient selections and a scalar inequality at every positive slice width, respectively, while (8) is stated for all probability measures supported in a fixed compact and their unique quadratic optimal maps from .