Discrete Optimization

Where MATH-style LPs let variables glide along a continuum, this graduate course is about what happens when decisions are integer or binary — assignments, on/off choices, routes — and the feasible region becomes a finite but exponentially large set you cannot just enumerate. You'll build IP and mixed-integer models for things like facility location and scheduling, then implement and benchmark the machinery that actually solves them: branch-and-bound, cutting planes, column generation, Lagrangian relaxation, and heuristics when exact methods give up. Two midterms and a final test the theory (polyhedral structure, complexity, duality), while a computational project pushes you to pick a real problem, design experiments, and defend your solution approach — the standard bridge between an LP background and OR research or operations roles where most interesting decisions are discrete.