A class of nonlinear optimisation and applications
This paper extends the theory on optimal transport, more precisely, as examples and applications in the final Section 5 reveal, transport problems from geometric optics in particular are to be covered. The known optimal transport theory -- ultimately going back to Monge and Kantorovich -- is connected to the field of partial differential equations via the Monge-Ampére equations, and thereby duality plays an important role. These connections are clearly exploited and extended in the structure of the present work and duality is given a dominant role. The intended extended construction of the transport optimization starts (in bounded regions \(U\) and \(V\)) with the definition of dual pairs \((u,v)\) of continuous functions. A given constraint \(\Phi\leq 0\) enters this definition. Under suitable conditions, using the implicit function theorem, a relation \(T:U\rightarrow V\) (and also its inverse \(T^{-1}\)) is obtained and then, after introducing measure-preserving mappings, the representation of a primal objective function \(I\) as a function to be maximized depending on \((u,v)\) and \(T\) follows. A dual task \(J(\mu)\rightarrow \min\) is formed (as usual) with the Lagrangian function, \(\mu\) is real because there is only one constraint. The tasks with \(I\) and \(J\) together with suitable conditions are cleverly formed, because a complete optimal transport theory can be derived, as a look at the remaining Sections 2--5 teaches. For maximizing of \(I\), there exist a maximizing pair \((u,v)\) (as in the earlier versions, \(u,v\) are called potentials) and an associated operator \(T\). The proof is given in Section 2 and contains esp. mollifications. Furthermore it is shown, that if the maximizing elements are smooth enough, say \(C^2\), then differentiation delivers, that they fullfil a Monge-Ampére-equation together with a boundary condition \((T(U)=V)\). The same is possible for the dual problem with \(J\) and the dual potential \(v\). In Section 3, generalized solutions are defined and it is shown that the maximizers defined above are also such solutions. Section 4 discusses Lagrange duality and it is especially proved that there is no duality gap. Section 5 contains several examples and interesting notes on optimization problems that can be treated with the presented theory.
- Convex duality in nonlinear optimal transport
- Constrained optimal transport
- A generalized dual maximizer for the Monge-Kantorovich transport problem
- A general duality theorem for the Monge-Kantorovich transport problem
- A simple relaxation approach to duality for optimal transport problems in completely regular spaces
- Convex duality in nonlinear optimal transport
- Dynamics of optimal partial transport
- scientific article; zbMATH DE number 5722561 (Why is no real title available?)
- scientific article; zbMATH DE number 19342 (Why is no real title available?)
- Class of optimization algorithms for the solution of large nonlinear programming problems. II
- scientific article; zbMATH DE number 7533281 (Why is no real title available?)
- Duality of Hessian manifolds and optimal transport
This page was built for publication: A class of nonlinear optimisation and applications
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q820868)
