Dynamic Programming and Stochastic Control
MIT OpenCourseWare presents this graduate-level course on sequential decision making under uncertainty. The syllabus covers optimal control of dynamical systems over finite and infinite horizons, including systems with finite or infinite state spaces and perfectly or imperfectly observed states. Lectures develop dynamic programming algorithms and approximation methods for problems where the state space is too large for exact solution. Recitations apply these techniques to problems drawn from engineering, economics, and operations research. Materials include lecture notes, assignments, and exams, following MIT's standard OCW format for its Electrical Engineering and Computer Science offerings. The course assumes familiarity with probability and linear algebra and builds toward the theory behind reinforcement learning and control algorithms used in robotics and finance. No certificate is offered, but all course materials are freely downloadable under MIT's open license.