Application of the Enhanced Semidefinite Relaxation Method to Construction of the Optimal Anisotropy Function

Abstract

In this paper we propose and apply the enhanced semidefinite relaxation technique for solving a class of non-convex quadratic optimization problems. The approach is based on enhancing the semidefinite relaxation methodology by complementing linear equality constraints by quadratic-linear constrains. We give sufficient conditions guaranteeing that the optimal values of the primal and enhanced semidefinite relaxed problems coincide. We apply this approach to the problem of resolving the optimal anisotropy function. The idea is to construct an optimal anisotropy function as a minimizer for the anisotropic interface energy functional for a given Jordan curve in the plane. We present computational examples of resolving the optimal anisotropy function. The examples include boundaries of real snowflakes.