Conforming approximation of convex functions with the finite element method (Q1679216)

Optimization problems with convexity constraints (OPCC) arise in many areas of research. When solving these problems, it is necessary to use efficient discretization strategies. In this paper, the approximation of convex functions by using the finite element method is considered. The author shows that all convex functions can be approximated by convex \({\mathcal P}^k\) functions for \(k\geq 2.\) This enables to consider a conforming discretization of OPCC. The convergence for \({\mathcal P}^k\) elements of interior approximations is shown are confirmed by numerical examples in the case \(k=2\). Similar results under additional constraints on the function values or on the gradient are provided. Numerical results are presented for computation of the projection of convex and non-convex functions on the set of convex functions and for the monopolist problem.