February 3, 2021
Professor Zelda Zabinsky – MW 8:30 – 9:50 am, Remote
Optimization is very useful in engineering design, operations, and strategic planning. Often these problems can be modeled with computer programs but lack a closed-form mathematical formulation. The problems often lack structure and may be non-convex, multi-modal, non-differentiable, and include both real-valued and integer variables. The problems may also include uncertainty.
The class will emphasize both formulating the problems as well as selecting appropriate solution techniques. We will discuss different ways to model uncertainty in the problem formulation and global optimization algorithms to solve the problems. Formulation issues will include ways to account for uncertainty, such as multiple objectives, CVaR, stochastic programs, and distributionally robust optimization. Global optimization methods will include deterministic methods such as interval methods and surrogate modeling and probabilistic search methods such as particle
filtering and Bayesian optimization.
Students will do readings and a project involving formulation and solution.
Prerequisite: Any prior exposure to optimization will satisfy the stated prerequisite of AMATH/MATH/INDE 515 and Math 328 or the instructor’s permission. Email zelda@uw.edu if you wish to take the course.