![[Dept. of Computer Science]](li-dsol.gif)
|
|
Jan. 11: The exam and final results after the retake of last friday are available (SOLIS-ID required).
Nov. 16: The exam and final results are available (SOLIS-ID required). For those with a final grade below 4: despite the formal rule, you are allowed to retake the exam in the 'herkansingsweek'.
Oct. 19: The results of the practical assignment are available (SOLIS-ID required).
Sept. 27:There are no more copies of the syllabus left at the student desk; I have therefore put a version online.
Sept. 16: The slides of the guest presentation are added. There are some Slovakian words in the presentation; here's their translation:
| ak, potom, inak | if, then, else | |
| a, alebo | and, or | |
| je | is (?) | |
| Obsluha, slabá, priemerná, výborná | Service, poor, good, excellent | |
| Jedlo, nechutné, chutné | Food, not tasty, tasty | |
| Prepitné, nízke, stredné, vysoké | Tip, low, medium, high | |
| agregovaná | aggregated |
Lecturer: dr. Silja Renooij
The following preliminary
table lists the meetings together with the topics for each lecture and
the corresponding syllabus chapters and course slides. In addition, the
relevant exercises per subject are listed.
| Week/day | Subjects | To study: (syllabus and corr. slides!) |
Exercises: (in syllabus) |
Slides [pdf] |
| 37 / Wed (Sept. 14, 2011) |
Introduction Probability and independence |
Ch. 1 +
Assignment Ch. 2 |
2.1 - 2.4 |
Ch1 Ch2+3 |
| 37 / Fri | Guest lecture by Alzbeta Michalíková on Fuzzy Sets | Fuzzy sets | ||
| 38 / Wed | Independence relations Undirected graphical models |
Ch. 3: § 3.1 Ch. 3: § 3.2.1 |
3.1 - 3.4 3.5, 3.6 |
Ch2+3 |
| 38 / Fri | Directed graphical models |
Ch. 3: § 3.2.2-3 |
3.7-3.14, |
" |
| 39 / Wed | Probabilistic network formalism Introduction to Pearl's algorithm |
Ch. 4: § 4.1 |
4.1, 4.2, 4.8 | Ch4 |
| 39 / Fri | NO CLASS! | |||
| 40 / Wed | Inference in singly connected graphs and trees |
Ch. 4: § 4.2.1, § 4.2.2 | 4.3 - 4.7 |
" |
| 40 / Fri | Examples of inference in singly conn. graphs Intro to inference in multiply connected graphs |
Ch. 4: § 4.2.3 |
4.9, 4.10, 4.13 |
" |
| 41 / Wed DEADLINE! |
Loop cutsets | " | 4.11,4.12 | |
| 41 / Fri | Construction of graph (by hand) Automated construction |
Ch. 5: §5.1, §5.2.1 Ch. 5: § 5.2.2 |
5.1 -5.3,5.4ab 5.7 |
Ch5 |
| 42 / Wed | Automated construction cntd Probability assessment |
Ch. 5: § 5.2.2 Ch. 5: § 5.3.1 |
5.7 | " |
| 42 / Fri | Noisy-or gate Expert assessment |
Ch. 5: § 5.3.2 Ch. 5: § 5.3.3 |
5.4cd, 5.5 5.6 |
" |
| 43 / Wed | Sensitivity Analysis | Ch. 5:
§ 5.3.4 Ch.6 : § 6.1 |
|
" Ch6 |
| 43 / Fri |
Sensitivity Analysis cntd Evaluation of PNs |
Ch. 6: § 6.2 |
6.1, 6.2 |
"
|
| 44 / Wed |
Evaluation and PNs as problem-solving architectures Last class: wrap-up + questions |
Ch. 6: § 6.3 |
" |
|
| 44 / Fri | No class! | |||
| 45 / Fri | Exam (see 'Tentamen' in the official schedule) | |||
| 1 / Fri | Re-Examinations ('aanvullende toets') |
The course will be graded based on a practical assignment and a written exam
FG = Mark(0.10 AG + 0.90 EG),
where Mark(x) =
Re-examination conditions:
To qualify for a second attempt at the written exam, the FG after the first
attempt at the written exam should
be at least a 4.0.
The practical assignment cannot be retaken.
Note that the re-examination week is week
1 of 2012 (see `Jaarindeling')!
This course is awarded 7.5 ECTS upon passing. It thus requires a total investment of approximately 210 hours, which means that apart from attending classes you are required to spend around 20 hours a week (!) on preparing and evaluating classes, working on the practical assignment, making exercises, and preparing for the exam.
The practical assigment consists of a number of questions and exercises to be answered by using one of the freely available
probabilistic network software tools. The associated question-answer form
should be handed in on paper no later than 16:59 hours on Wednesday the 12th of October
(during class, or in the pigeon hole of the lecturer (BBL 513));
this deadline will be strictly
enforced!
To pass the written exam, an in-depth understanding of and insight in the subjects treated is required.
Details of what is expected of you can be found in the studymanual[pdf].
Two examples of previous exams (in dutch; there is a brief english translation[txt]) are
available (exam1.pdf, exam2.pdf).
Syllabus: Probabilistic Reasoning, authors: L.C. van der Gaag and S. Renooij. (version history)
Studymanual: Studymanual Probabilistic Reasoning, author: S. Renooij. (version history)
Course slides: although the syllabus covers all subjects, some are discussed in more detail in the course slides.
You should be familiar with these additional details. (version history)
Further reading:
For those seeking additional material, I recommend the following:
- recent textbooks on probabilistic networks (with solutions to lots of exercises on books' own websites):
(Note: online versions of both books are available through SpringerLink)
- Bayesian Networks and Decision Graphs, by Finn V. Jensen and Thomas D. Nielsen (2007)
- Bayesian Networks and Influence Diagrams: A Guide to Construction and Analysis, by Uffe B. Kjærulff and Anders L Madsen (2008)
- online tutorials and background info:
- The Helsinki B-Course;
- The Monash university Bayesian Artificial Intelligence page (book, tutorials, projects, talks);
- various resources listed by Agena Risk;
- video lectures, search on e.g. videolectures.net, vidoemo.com, ...
- on probability theory and statistics: the Virtual Laboratories in Probability and Statistics
- on d-separation (history and explanation): D-separation
- on independence relations and Pearl's algorithm:
J. Pearl
Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference
- on most subjects concerning the construction and theory of probabilistic networks: DSS-publications
The software listed below may help you to gain more insight in inference in (Chapter 4) and the construction of (Chapter 5) probabilistic networks. Numerous software packages have been developed for constructing and reasoning with probabilistic networks; most of them have a free downloadable (demo-) version. Install the one of your choice for the practical assignment. (I recommend Genie or SamIam, or if you prefer a commercial product: Hugin-Lite or Netica.)Applications
- University packages (free; some require registration):
- BayesBuilder (Universiteit van Nijmegen; Platform: Windows; Documentation: online help)
- Dazzle (Universiteit Utrecht; Platform: Windows; Documentation: non-existent)
- Elvira (8 Universities in Spain; Platform: MS-DOS/Windows, Linux, Solaris; Documentation: online, in Spanish)
- Genie (Pittsburgh University; Platform: Windows, Unix (Solaris), Linux, Mac, Pocket PC; Documentation: online; registration required)
- SamIam (UCLA; Platform: Windows, Solaris, Linux; Documentation: online; registration required)
- Commercial Packages (free limited versions; some require registration):
- Hugin -Lite (HUGIN EXPERT A/S; Platform: Windows, Solaris, Linux, Mac OS; Documentation: online; requires registration)
- MSBNx (Microsoft Research; Platform: Windows; Documentation: online)
- Netica (Norsys Software Corp.; Platform: Windows, Mac OS; Documentation: online tutorial)
- Overviews of packages
- overview of software for probabilistic networks, including example networks and datasets;
- another overview
- Some packages have downloadable as well as online tools (applets) (drawback: networks cannot be saved):
- Java Bayes (University of Sao Paulo);
- Bayes Applet (University of British Columbia; Belief and Decision Networks);
Applications of Probabilistic Networks exist for several different problem domains. The following list of references point to a collection of well-known networks in different formats (some are toy-examples!), and some papers about different networks constructed in our Department:
the Bayesian network repository and other (links to) network collections;
Bayesian networks: a practical guide to applications;
network for information disclosure[ps];
VSD network [pdf];
the oesophageal cancer network[pdf];
network for classical swine fever [pdf];
- The MSc programme `Applied Computing Science' (continued in the new programme 'Computing Science' as of Sept. 2010);
- The MSc programme `Computing Science';
- The MSc programme `Technical Artificial Intelligence';
- The Decision Support Systems group, and its Experimentation and Graduation Projects
- The Algorithmic Systems group
- The Algorithmic Data Analysis group
- The Software Technology group
Course evaluation
To guard the quality of our educational programme, each year every course is evaluated by the students that registered for the course. The student-evaluation helps instructors to evaluate and if necessary adapt their courses and allows the Institute to evaluate its teachers. For the evaluations to be useful it is necessary to get respons from a large number of participants.My intention is that you find the course on probabilistic reasoning interesting, motivating, perhaps a bit difficult now and again but doable, challenging and enjoyable, and last but not least I hope that you feel that you have learnt many useful things. If for some reason the course wasn't what you'd expected and/or if you have any (reasonable) suggestions to improve the course and everything that comes with it, please fill out the evaluation form. Suggestions from the evaluations have for example resulted in online and numbered transparancies. If you're completely satisfied then I'd also like to know that and filling out the evaluation form would be a way of telling me this.
For those who are interested, I provide links to the evaluations of previous years. Note that these can be viewed internally only. For each evaluation I list the number of registrants (union of OSIRIS subscribers and participants), the number of participants (distinct students that participated in and submitted at least one exam in that year), and the number of respondents (students that filled out the evaluation form).
Previous lecturer:
Current lecturer:
- 1999/2000 (page 37) (respondents: 9)
- 2000/2001 (33 registrants, 24 participants, 5 respondents)
- 2001/2002 (30 registrants, 22 participants, 7 respondents)
- 2002/2003 (40 registrants, 33 participants, 16 respondents)
- 2003/2004 (37 registrants, 31 participants, 10 respondents)
- 2004/2005; period 1 "Probabilistisch Redeneren" (26 registrants, 21 participants, 10 respondents)
- 2005/2006; period 1 "Probabilistisch Redeneren" (33 registrants, 28 participants, 16 respondents)
- 2006/2007; period 1 "Probabilistisch Redeneren" (20 registrants, 24 participants, 10 respondents)
- 2007/2008; period 1 "Probabilistisch Redeneren" (25 registrants, 23 participants, 12 respondents)
- 2008/2009; period 1 "Probabilistisch Redeneren" (23 registrants, 21 participants, 8 respondents)
- 2009/2010; period 1 "Probabilistisch Redeneren" (24 registrants, 21 participants, 13 respondents)
- 2010/2011; period 1 "Probabilistisch Redeneren" (25 registrants, 23 participants, 10 respondents)
- 2011/2012; period 1 "Probabilistisch Redeneren" (33 registrants, ?? participants, ?? respondents)
Silja Renooij November, 2011.