A measure-valued approach to convex set-valued dynamics (Q5959755)

A new approach to the evolution of convex-valued multifunctions is presented, analogous to single-valued gradient dynamics. Simple convex-valued multifunctions are introduced to approximate the solution of a convex-valued differential equation; then the differentiability of these multifunctions is investigated. A constraint stochastic optimization problem is also presented. The approach to set-valued differentiation developed in this paper essentially differs from existing ones and thus may be a fruitful source of open problems for further research.