|Website:||website containing additional information|
|Period:||period 1 (week 36 through 45, i.e., 2-9-2019 through 8-11-2019; retake week 2)|
|Participants:||see Osiris Docent|
|Schedule:||Official schedule representation can be found in Osiris|
|Contents:||In this course, we will consider the theory and applicability of Bayesian
networks, which are a member of the family of Probabilistic Graphical Models (PGM). See the Description below on this page to get a flavour of their use.|
This course roughly consists of three parts. As a general introduction to (probabilistic) graphical models, the first part of the course deals with independence relations and their representation by means of undirected and directed graphs. The second part introduces the Bayesian network as a compact representation of a probability distribution on a set of statistical variables; in addition, algorithms that allow for efficiently computing probabilities from a Bayesian network are discussed. The third part of the course concerns the construction of Bayesian networks for real-life applications. Topics covered include automated construction of networks from data, handcrafting the network with the help of domain experts, robustness analysis and evaluation.
Learning goals (2019-2020): upon completing this course, the student
1. recognises and understands the strengths and weaknesses of probabilistic graphical models in general and Bayesian networks in particular;
2. understands the relation between probabilistic independence and the graphical representations thereof, and is able to draw conclusions from this relation;
3. understands and is able to apply probabilistic inference in Bayesian networks;
4. has knowledge and understanding of methods for constructing the Bayesian network graph for actual applications;
5. has knowledge and understanding of various methods for quantifying Bayesian networks, including parameterized models (noisy-or), together with their benefits and limitations;
6. understands and is able to apply techniques for evaluating the robustness and quality of a Bayesian network.
Warning: To be able to fulfill the learning goals of this course, a solid understanding of the mathematics underlying PGMs is required. This means you should have sufficient mathematical skills to understand, apply and manipulate the many formulas that are presented. (You can have a look at the course slides to get an impression of the level of mathematics involved.) In addition, it is assumed that you are capable of abstracting away from given examples, applying the knowledge and techniques learned to contexts other than those discussed in class.
1. Syllabus 'Probabilistic Reasoning': paper edition sold by A-Eskwadraat; up-to-date online edition available through the website with additional information
2. Course slides, available online.
Additional: Studymanual, also available online.
|Course form:||- Lectures (twice a week)|
- Self-assessment exercises
|Exam form:||A set of practical homework assignments (15% in total) and one written test (85%). If you are registered for the course, you are automatically entitled to partaking in these assessments. Those who are allowed to do a substitute test (see re-examination conditions) are automatically registered for this. |
Note that course registration proceeds through OSIRIS and generally closes about 8 weeks before the start of the course (see registration dates)! If you are planning on taking this course in your first year, you will be notified how to register during the Master Introduction. The lecturer cannot do the course registration for you.
|Minimum effort to qualify for 2nd chance exam:||Dictated by Graduate School: you have failed the course with a final grade of at least 4 after the first attempt. |
Additional conditions specific to this course: see Examination and Grading page on website with additional information.
|Description:||Bayesian networks can be used for reasoning and decision support under uncertainty:
Which exercises are most suitable for Bob to improve his calculus skills?
In complex domains, people have to make judgments and decisions based on uncertain, and often even conflicting, information; a difficult task, even for experts in the domain. To support these complex decisions, knowledge-based systems should be able to cope with this type of information. For this reason, models for representing uncertainty and algorithms for manipulating uncertain information are important research subjects within the field of Artificial Intelligence. Probability theory is one of the oldest theories dealing with the concept of uncertainty and therefore plays an important role in many decision support systems.
In this course, we will consider probabilistic models for representing and reasoning under uncertainty. More specifically, we will consider the theory underlying the framework of Bayesian networks, their definition and reasoning, and discuss issues and methods related to the construction of such networks for real-life applications.