On Monge-Kantorovich problem in the plane (Q960991): Difference between revisions

Optimal transport plans in Euclidean space have been characterized in terms of gradients of convex functions in Brenier (1987, 1991) and in Rüschendorf and Rachev (1990). In the one-dimensional case optimal transforms are known in explicit form and are given by inverse distribution functions. The aim of the authors is to make this idea also fruitful in the two-dimensional case. If \((X_1, X_2)\), \((Y_1, Y_2)\) is an optimal coupling of probability measures \(P\) and \(Q\), then the authors construct this optimal coupling as functional of \(Z=(X_1, Y_2)\) given by conditional distribution functions of the joint distribution function \(H\) of \(Z\). They show that the optimal \(H\) and thus the optimal coupling is determined by an quasi-elliptic equation.