Derivative Free Optimization

This course is aimed for the target audience of graduate students. Several engineering and scientific problems lead to optimization settings, where the objectives and constraints are governed by complex and time-consuming simulations. Additionally, there exist no-closed form expressions for these functions (objectives and/or constraints). Further, several of these simulators are proprietary, thus making applicability of automatic differentiation less conductive. The last two decades have made significant inroads by looking at either model based methods or direct search techniques to solve such problems. More importantly, both these classes of algorithms have established convergence theory under some mild problem specific assumptions. This course focuses on theory, algorithms, and applications.