Title: Mesh Adaptive Direct Search for Nonlinear and Mixed Variable Constrained Optimization
Abstract: In this talk, we present the class of mesh adaptive direct search (MADS) algorithms for solving very general constrained optimization problems, in which variables may be continuous or categorical, and objective and constraint functions may be discontinuous, nonsmooth, nonconvex, computational costly to evaluate, or fail unexpectedly. This class of problems is common in engineering problems, especially when function evaluations require the running of an engineering simulation code. Convergence to limit points satisfying suitably defined optimality conditions will be discussed, and three applications are described, in which MADS was applied to achieve successful results.