January 13 (Tue) and 15 (Thu), 1998 Introduction and Overview [S]: Building a KBS. Rule-Based Systems: Chpater 1 [J]
January 20 and 22 Chapter 2 [J]: Introduction to Bayesian Networks. Homework Assigned: Exercises 2.1-2.4 (8/20, collected 8/22) and 2.5-2.6 (8/22, to be collected 8/29).
January 27 Definition of Bayesian Net; The Chain Rule; Numerical Examples (Ch.2 [J]). Building Models: Catching the Structure: The Family Out and Insemination Examples; Naive Bayes Models [J, 3.1]. Homework Assigned: All remaining exercises in Ch.2 (through 2.9).
January 29 (Thursday) Review of Icy Roads Example, with use of clusters; Determining the Conditional Probabilities [J, 3.2]; Some Modeling Tricks: Constraints and Noisy OR [J, 3.3.1, 3.3.2]. Homework Assigned: Exercises 3.2, 3.8, 3.13 [J].
February 3 (Tuesday) Correction of Exercise 3.2, 3.8, 3.13. More modeling tricks [J, 3,3]. Exercise 3.4 assigned, due on February 12.
February 5 (Thursday) Propagation in Bayesian Networks: up to Theorem 4.3 included. No new homework assigned.
February 10 (Tuesday) Propagation in Bayesian Networks: Trees. Homework Assigned: Exercise 3.15 (Scene Interpretation) assigned, due on 2/17.
February 12 (Thursday) Propagation in Bayesian Networks: Chapter completed, except for Stochastic Simulation. Homework Assigned: Exercises 4.6, 4.8 (part iv only), 4.9, 4.11 (parts i-iii only), due 2/17.
February 17 (Tuesday)
February 19 (Thursday)
February 24 (Tuesday)
February 26 (Thursday) Non-Serial DP and Its Relation to Probability Propagation [Notes]. Homework Assigned: 5.13 p. 131, 6.1(i only) p.154 [J] all due 3/3. 6.2 and 6.3 p.154 [J] using Hugin, due 3/17 (after Spring break)
March 3 (Tuesday) Discussion of the insemination network and related exercise (5.13). Exercises 5.13 and 6.1 (i only) [J] are now due on 3/5. Symbol Systems, ch. 1 [S] (all); discussion of assigned exercises (1.1.5, 1.1.6, 1.4.2, 1.4.5, 1.4.7 [S], due 3/5.
March 5 (Thursday) The complexity of terminological reasoning, and especially of subsumption. References: 1) Nebel, Bernhard. "Computational Complexity of Terminological Reasoning in BACK." Artificial Intelligence, 35 (1988). 371-383. 2) Brachman, R.J. and H.J. Levesque. "The Tractability of Subsumption in Frame-Based Description Languages." Proceedings of AAAI-94 (Austin, TX, 1994), pp.34-37.
March 8-16: Spring Break
March 17 (Tuesday) Exercises 1.1.5, 1.1.6, 1.4.2, 1.4.5, 1.4.7 [S] collected. Hugin Programming Exercises 6.2 (selection of golf implement influence diagram) and 6.3 (breeding) will be demonstrated today, Wednesday, and Friday, individually, by each student. Knowledge Acquisition: 3.1 and 3.2 [S].
March 19 (Thursday) Exercises 3.2.9, 3.3.5, and 3.3.7 assigned (due Tuesday, March 23). Knowledge Acquisition and MYCIN: 3.3 and 3.4 [S], with supplements on KADS, especially examples of interpretation models, such as for assessment. Paper handed out: Karbach, W., M. Linster, and A. Voss. "Models, Methods, Roles and Tasks: Many Labels--One Idea?" Knowledge Acquisition, 2 (1990), 279-299.
March 24 (Tuesday) Exercise 3.15H [J] (Scene interpretation using Bayesian networks) assigned. Results will be demonstrated on Friday, March 27. Classification: Section 7.1 and 7.2 [S]. Exercises 7.2.1, 7.2.4, and 7.2.6 carried out in class. I used HyTEx (Hybrid Toxicology Expert System) as an example of a system that uses classification as an initial task, and less analytic tasks (e.g., prediction) later on. I also used the examples of heuristic classification/heuristic diagnosis in the KADS IM deliverable.
March 26 (Thursday) Classification in MYCIN, MORE, MOLE, and MDX (7.3 and 7.4). Begin discussion of model-based diagnosis.
March 31 (Tuesday) Discussion of Raymond Reiter's "A Theory of Diagnosis from First Principles," Artificial Intelligence 32, 1987, 57-95. Students are asked to bring in a typed one-page proposal for a Bayesian network based KBS, extensible to an influence diagram, on Thursday, April 2.
April 2 (Thursday)