Duality Relations and Numerical Methods for Optimization Problems on the Space of Probability Measures with Constraints on Probability Densities
Abstract
In this paper, the authors look at some quite general optimization problems on the space of probabilistic measures. These problems originated in mathematical statistics but have applications in several other areas of mathematical analysis. The authors extend previous work by considering a more general form of the constraints, and develop numerical methods (based on stochastic quasigradient techniques) and some duality relations for problems of this type.
This paper is a contribution to research on stochastic optimization currently underway within the Adaptation and Optimization Project.