CMSC5728: Decision Analysis and Game Theory
General Expectations:
Student/Faculty Expectations on Teaching and Learning
Teacher:
Prof. John C.S. Lui
This is a graduate level course which covers theory on decision science.
There are several main topics I plan to cover, they are:
(a) Multiarmed bandit theory;
(b) Game theory;
(c) Reinforcement learning theory.
I like to emphasize that course
is mathematical and algorithmic in nature.
I will introduce a lot of concepts, show the mathematical proves,
and present the physical meanings and applications.
Students are expected to follow and understand my lecture,
and also do a lot of readings and do some programming (via Python).
Important reminder:
Students
are expected to attend the lecutre,
read the leture notes and understand them,
spend time to read resources on the Internet,
do the homework,
do the programming assignments,..etc,
so to keep pace with this course.
Teaching Assistants
Reference:
Course Grades:
 Written Homework and/or Programming Assignment: 50%
 Exam: 50%
Important note:
Students need to get at least 30% in the final exam to pass, independent
of their performance in programming exercises.
IMPORTANT REMINDERS !!!!!!
 Final Examination will be on December XXX, 2020. X:XX pm till Y:YY pm.
Venue will be XXX YYY.
Policies:
 No late homework, programming assignments or projects will be accepted;
Outline for the course:
(Note: I usually prepare more materials
than we can cover in a semester. I will leave those materials I can't
cover to students as a selflearning tool.)
 Introduction to topics on decision science
 Introduction to Game theory
 Twoplayer game zerosum games
 Dominance stratey
 Saddle point
 Mixed strategy
 Minimax theorem
 Twoplayer game nonzero sum games
 Concept of equilibrium (Nash Equilibrium)
 Cournot Model of Duopoly
 Dynamic Games
 Kuhn's Theorem
 Concept of Subgame
 Subgame Perfect Nash Equilibrium
 Games with continuous strategy space
 Stackelberg Games
 Introduciton to Coalition and Cooperative Games
 Auctions
 Mechanism Design
 Stochastic multiarmed bandit (MAB)
 UCB algorithms and regret bound
 Thompson Sampling and its application to MAB
 Adersarial Bandits
 Linear Bandits
 Contextual Bandit
 MAB application: Dynamic Pricing, networking, crowdsourcing and multipath protocols
 Markov Decision Process
 ...etc
Lecture Notes
(Lecture Notes are available at CUHK Blackboard (https://blackboard.cuhk.edu.hk/))

Please refer to the CUHK Blackboard
Written homework and programming assignment

Please go to the "Blackboard" to access the specification.