### Academic year 2019/2020:

**October 3, 2019:** I was asked some questions about assignment D in class, so apparently the instructions as to what you can or cannot compute using your software are not completely clear. First of all: the Car diagnosis network is too large an example to do Pearl (with loopcutset conditioning!) by hand, so that is not required of you. You are allowed to, and supposed to, use your software to compute distributions of the form Pr(V_i), or Pr(V_i | evidence), that is (conditional) probability distributions over a *single* variable. More complex queries first have to be rewritten in terms of distributions of the above-mentioned form, even if your software is capable of directly answering the more complex query for you (you can then use that to check your answer).

*Learning goals (2019-2020):* upon completing this course, the student

- recognises and understands the strengths and weaknesses of probabilistic graphical models in general and Bayesian networks in particular;
- understands the relation between probabilistic independence and the graphical representations thereof, and is able to draw conclusions from this relation;
- understands and is able to apply probabilistic inference in Bayesian networks;
- has knowledge and understanding of methods for constructing the Bayesian network graph for actual applications;
- has knowledge and understanding of various methods for quantifying Bayesian networks, including parameterized models (noisy-or), together with their benefits and limitations;
- understands and is able to apply techniques for evaluating the robustness and quality of a Bayesian network.