Course Objective

After completing this course, the student can
1. understand the applicability and the limitations of discrete- and
continuous-time Markov decision processes
2. can model real-world decision problems into the Markov decision
theoretic framework
3. can implement decision models to calculate the optimal decision rules
4. can mathematically prove properties of the optimal decision rules in
specific systems
5. can deal with modeling and implementation issues related to the curse
of dimensionality
6. can model problems with partial observations
​

Course Content

This course deals with the theory and algorithms for stochastic
optimization with an application to controlled stochastic systems (e.g.,
call center management, inventory control, optimal design of
communication networks). We discuss aspects of semi-Markov decision
theory and their applications in certain queueing systems and also
discuss parts of the related field reinforcement learning. In
programming assignments, students learn to implement optimization
algorithms and experiment with them. Programming is done in R.

Teaching Methods

Lectures.

Method of Assessment

Programming exercises, final exam.
Assignments: 50%
Exam: 50%
The weighted average needs to be 5.5 or higher.

Entry Requirements

Programming experience

Literature

Lecture notes will be posted on Canvas.

Target Audience

mBA, mBa-D, mMath, mSFM.

Recommended background knowledge

Stochastic Modeling (X_400646) or equivalent courses on Stochastic
Processes and Queueing Theory.