The Setting
Moving the atoms of a target measure at fixed masses changes the intercepts of its finite affine potential. The estimates concern potential differences along such paths and the resulting stability of optimal transport maps.
Let be compact and convex with nonempty interior. The uniform measure on is normalized Lebesgue measure, as in (1); let bound for -almost every .✦ (1)✦
For a finite nonempty index set , sites and intercepts , the finite affine potential is the function in (2). The closed Laguerre cell is the subset of where the th affine function attains the maximum, including ties; its normalized cell mass is its -mass, as in (3). (2)✦ (3)
For weights , the weighted intercept mean is the linear functional in (4); its zero-mean hyperplane is the subspace on which it vanishes. For positive masses with , the canonical normalized intercept vector is a vector in assigning the prescribed masses at distinct sites and equal to zero when sites repeat, as in (5). (4) (5)
A supporting pair on consists of a function and a subgradient selection satisfying (6). For , a supporting pair on is -bounded when is measurable and everywhere. (6)✦
A coupling of probabilities on is a probability measure on with these two marginals; write for their couplings.✦ The graph transport plan is the pushforward measure in (7), and is the quadratic cost functional of a plan. For measures on , the quadratic Wasserstein distance is the square root of the real value of the infimum of their attainable quadratic coupling costs. (7)
For a probability on , a quadratic optimal map from to is a measurable map with whose graph plan minimizes quadratic cost among couplings.✦ Uniqueness of a quadratic optimal map from to means that every coupling of no greater cost equals ; denote such a map by .✦ A measure on is carried by a set when it assigns zero mass to .✦ For maps , the map distance is the square root of .✦
For site and intercept velocities , the energy bounds with parameters mean (8) at every . (8)
The Claim
Theorem 1. Let be a second family of sites in . Suppose , , and the linear site path has distinct sites for . Assume its canonical intercepts are differentiable there and satisfy (8) with and . For , the interior estimate (9) holds whenever : (9)✦
For the endpoint assertion, may be any set whose uniform measure is finite and satisfies the same almost everywhere norm bound. Let . Suppose the site paths are continuous, with distinct sites on , and let be intercept paths. Assume , on , where are continuous on . Assume (8) and on . Suppose a measurable selector attains the affine maximum for every at each interior time, and there. There exist intercepts whose potentials satisfy (10): (10)✦
Let again be the compact convex source body. Retain the endpoint path hypotheses, replacing the selector assumptions by measurability of each and for . Assume also and . There are -bounded supporting pairs with the endpoint properties in (11): (11)✦
Let be compact and nonempty, with on , and let . Suppose every finite positive mass vector of total mass one and every pair of site families satisfying , whose linear interpolation has distinct sites at every interior time, admit -bounded supporting pairs sending to and , respectively, and a real constant satisfying (12): (12) There is such that every pair of probabilities carried by has unique quadratic optimal maps from obeying (13): (13)✦
In the paper, the result reads: for the uniform source on and any finitely supported probabilities in , there are convex -Lipschitz finite maxima of affine functions, with and -almost everywhere, such that .
What Is Checked
The estimates (9), (10) and (11) are stated over finite moving-site paths and normalized Lebesgue measure, while (13) quantifies over all probabilities carried by the compact target container . For the source body and distinct sites, the normalized intercept vector in (5) assigning prescribed positive masses of total mass one exists uniquely.✦