The Setting
Let be a convex function on whose gradient map is a homeomorphism of onto itself and pushes the source measure forward to a constant multiple of the target measure , where is a real function on , the potential, and . This is the moment measure picture behind the Monge-Ampère equation , where is a real constant. Source coordinates carry , and target coordinates carry .
Weighted integration by parts under happens in , while and its Hessian are known at . The objects that pass between the two are the Hessian of read as a function of , the second-order operator it defines on test functions of , and the flux of a test function, obtained by applying to its gradient. For a test function there are also a matrix field and a vector field , the weighted divergence of under ; both are built in source coordinates, and a function of source coordinates is read in target coordinates by composing it with the inverse of the gradient map.
Coordinates, Derivatives and Weights
The coordinate space of a finite index set is the real vector space of all functions from to : a point has one real coordinate for each index, and points are added and scaled coordinatewise.✦ It is when has elements. The standard basis vector of has coordinate at the index and at every other index.✦
The directional derivative of a function on a real normed space along a vector of is the function , where is the Fréchet derivative of at .✦ On , the partial derivative is .✦ The gradient of a function on at a point is the vector of its partial derivatives, and its Hessian is the real matrix of its second partial derivatives: (1)✦ (2)✦
The weight of a potential on is the positive function .✦ For a vector of , the weighted adjoint of for this weight sends a function on to .✦ The weighted divergence of a family of functions on , along a family of vectors of , is the function : (3)✦
Fix a measure on ; in the setting it is Lebesgue measure. The source measure of a function on is the measure with density with respect to , and the target measure of a potential and a constant is the measure with density with respect to : (4)✦ (5)✦
The Definition
Definition 1. Let and be real functions on , and let be a homeomorphism of onto itself; in the setting, is the gradient map . The target Hessian of is the matrix field (6)✦ The moment operator sends to the function (7)✦ where . The flux of is the vector field (8)✦ In source coordinates, for indices and , let be the entries of the product of with the Hessian of pulled back by , and let .✦ The weighted divergence of for the potential , taken as in (3) along the standard basis, is (9)✦ Let .✦ A function of source coordinates is read in target coordinates as ✦. The matrix field is read in target coordinates, and the vector field has as components the functions read in target coordinates: (10)✦ (11)✦
What Is Checked
The definitions are stated over the coordinate space of a finite index set , with the weighted adjoint and the weighted divergence over any real normed space and every derivative a Fréchet derivative, and their data are a homeomorphism of onto itself, real functions , and , a real constant and a measure on .