Important Notice
Due to unforeseen circumstances, the Bayesian Networks course will not be given in Spring semester 2017-2018. The course will be replaced by the course Reinforcement Learning. All students who have registered for the Bayesian Networks course are advised to follow the Reinforcement Learning course.
Admission requirements
The student should have taken a course in AI.
Description
In 2012 Judea Pearl was given the Turing Award (seen as the Nobel prize in computer science) for his groundbreaking work on probabilistic and causal reasoning in intelligent systems. His work forms the core of this artificial intelligence (AI) course, which now, because of the Turing Award, is generally seen as one of the most important current topics of computer science. Handling uncertain knowledge has been one of the central problems of AI research during the past 30 years. In the 1970s and 1980s uncertainty was handled by means of formalisms that were linked to rule-based representation and reasoning methods. Since the 1990s probabilistic graphical models, in particular Bayesian networks, are seen as the primary formalisms to deal with uncertain knowledge. Various fundamental aspects of probabilistic graphical models, such as graphical representation, the logic of independence, reasoning and learning, are studied in this course, with some emphasis on Bayesian networks. In addition, building Bayesian networks for real-world applications will be covered and students will obtain some experience in how to build Bayesian networks for a problem domain in the practical and through small projects. Connections to cognitive science can be established in the associated seminar.
Course objectives
At the end of the Bayesian Network course, the student should be able to:
understand the probabilistic principles of reasoning under uncertainty
explain the differences between various graphical models in particular in terms of representation of independence information
have insight into algorithms for probabilistic reasoning in Bayesian networks
have insight into the pros and cons of learning models versus using expert knowledge
have some experience in exploiting software to solve problems involving uncertainty
in the seminar, student will learn how to interpreter the scientific literature on Bayesian networks and related probabilistic graphical models.
Timetable
The most recent timetable can be found at the LIACS website.
Mode of instruction
lectures
seminar
tutorials
practical assignment/project
Assessment method
The final mark is composed of (1) written exam (60%) (2) practical assignment (40%) (3) seminar and essay (optional, then: written exam 40%, assignments 30%, seminar 30%)
Reading list
Mandatory:
K.B. Korb and A.E. Nicholson, Bayesian Artificial Intelligence, Chapman & Hall, Boca Raton, 2004 or 2010
Background literature:
R.G. Cowell, A.P. Dawid, S.L. Lauritzen and D.J. Spiegelhalter, Probabilistic Networks and Expert Systems, Springer, New York, 1999
F.V. Jensen and T. Nielsen, Bayesian Networks and Decision Graphs, Springer, New York, 2007
D. Koller and N. Friedman, Probabilistic Graphical Models: Principles and Techniques, MIT Press, Cambridge, MA, 2009
Signing up for classes and exams
You can enrol via uSis . More information about signing up for classes and exams can be found here .
There is limited space for students who are not enrolled in the BSc programme of Computer Science or the Minor Data Science. Please contact the study coordinator/study adviser.
Contact information
Programme coordinator: Ms. José Visser.