Novel Techniques for Mathematical Optimization of Black-Box Nonlinear Functions

We have developed and tested a number of novel techniques for mathematical optimization of black-box multidimensional nonlinear real-valued continuous functions, with focus on scenarios where the predominant cost of optimization is the time complexity due to the number of function evaluations, such as is commonly the case with the Princeton Iterative Equilibrium Solver (PIES) code for MHD equilibria calculation. Our research builds on the existing Jacobian-free Newton-Krylov (JFNK) framework. We present a review of existing techniques in addition to our results and analysis.