CSCE 790 (Spring 2019) Lecture Log

January 15 (Tue), 2019 Introduction to the Course: Goals and Topics to Be Covered. Students introduce themselves. Hugin. Examples of plausible reasoning and causal graphs: icy roads (started).

January 17 (Thu), 2019 Detailed computation (by hand) for the icy roads example with review of basic probability concepts. Findings and evidence. Probability of the evidence.

January 22 (Tue), 2019 More examples. Problems with compositional systems. Induced dependencies. Rumors. Kolmogorov's axioms. Classical model of probability. Subjective (Bayesian) model, started.

January 24 (Thu), 2019 Students are asked to memorize the "Visit to Asia" (also known as "Asia" and "Chest Clinic") network, which is given in the introductory slides. The following two papers are assigned for presentation on Thursday, January 31:

  1. Rina Dechter. "Probabilistic Networks." Chapter 14 in Constraint Processing. San Francisco: Morgan Kaufmann, 2003. (To be presented by Noah Geveke.)
  2. Steffen L. Lauritzen and Finn V. Jensen, "Local Computations with Valuations from a Commutative Semigroup," Annals of Mathematics and Artificial Intelligence, 21 (1997), pp.51-69. (To be presented by Rui Xin.)
The subjective (Bayesian) interpretation of probability is a model of Kolmogorov axioms. The directed local Markov condition and the definition of Bayesian network (according to [N90]; other definitions, which are almost always equivalent, exist). The chain rule for Bayesian networks. Any (discrete, finite) Bayesian network is identified by its DAG and the conditional probability table for each of its nodes.

January 29 (Tue), 2019 Discussion of presentation format. Guidelines and examples are on the course website. The Chernobyl example. Slides and notes used in the spring 2016 offering of CSCE 582 are now linked to the course website. We go over a few of them, including the notes that have a proof of the chain rule for Bayesian networks. The global Markov condition; d-separation can be computed using the global Markov condition. Causality: the coins and bells example started.

January 31 (Thu), 2019 Noah Geveke presents the following paper: Rina Dechter. "Probabilistic Networks." Chapter 14 in Constraint Processing. San Francisco: Morgan Kaufmann, 2003. Clairvoyance and the definition of variables in Bayesian networks. Additional detailed example of the bucket elimination algorithm: computing the probability of tuberculosis given a visit to Asia and dyspnea in the Asia (chest clinic) example.

February 5 (Tue), 2019 The Junction Tree algorithm (Finn Verner Jensen's Hugin version). Lecture based on ch.4 [J96], through section 4.3; to be continued on Thursday.

February 7 (Thu), 2019 The Junction Tree algorithm, continued through Section 4.5.1.

February 12 (Tue), 2019 Homework (HW1) assigned: due all exercises in Ch.4 [J96], due by Tuesday, February 26. The Junction Tree algorithm, completed (i.e., Ch.4 [J96] through Section 4.5).

February 14 (Thu), 2019 Rui Xin presents the following paper: Steffen L. Lauritzen and Finn V. Jensen, "Local Computations with Valuations from a Commutative Semigroup," Annals of Mathematics and Artificial Intelligence, 21 (1997), pp.51-69.
The following papers are assigned for presentation

  1. Finn Verner Jensen and Frank Jensen. "Optimal Junction Trees." Proceedings of the Tenth Conference on Uncertainty in Artificial Intelligence (UAI-1994), pp.360-366. (To be presented by Nattapon Donratanapat on Thursday, February 21. Update of 2019-02-19: canceled, to be reassigned.)
  2. Anders L. Madsen and Finn Verner Jensen. "Lazy Propagation in Junction Trees." In: G.F. Cooper and S. Moral (eds.), Uncertainty in Artificial Intelligence: Proceedings of the Fourteenth Conference. San Francisco, CA: Morgan-Kaufmann, 1998, pp.362-369. (To be presented by MD Shahriar Iqbal on Thursday, February 21.)
  3. Mark Bloemeke and Marco Valtorta. "A Hybrid Algorithm to Compute Marginal and Joint Beliefs in Bayesian Networks and Its Complexity." In: G.F. Cooper and S. Moral (eds.), Uncertainty in Artificial Intelligence: Proceedings of the Fourteenth Conference. San Francisco, CA: Morgan-Kaufmann, 1998, pp.16-23. (To be presented by Sun on Tuesday, February 26.)
  4. Anders Madsen, Cory Butz, Jhonathan Oliveira, and Andre Dos Santos. "Simple Propagation with Arc Reversal in Bayesian Networks." Proceedings of the 9th International Conference on Probabilistic Graphical Models (PGM-18), PMLR-72 (Vaclav Kratochvil, Milan Studeny, editors), 2018, pp.260-271. (To be presented by Jianhai Su on Tuesday, February 26.)

February 19 (Tue), 2019 Variable elimination revisited: Non-Serial Dynamic Programming (NSDP). Variable elimination in constraint satisfaction problems (CSPs): a relational algebra perspecives, following the presentation by Poole and Mackworth. The following paper was re-assigned for presentation

  1. Finn Verner Jensen and Frank Jensen. "Optimal Junction Trees." Proceedings of the Tenth Conference on Uncertainty in Artificial Intelligence (UAI-1994), pp.360-366. (To be presented by Noah Geveke on Tuesday, February 26.)

February 21 (Thu), 2019 MD Shahriar Iqbal presents: Anders L. Madsen and Finn Verner Jensen. "Lazy Propagation in Junction Trees." In: G.F. Cooper and S. Moral (eds.), Uncertainty in Artificial Intelligence: Proceedings of the Fourteenth Conference. San Francisco, CA: Morgan-Kaufmann, 1998, pp.362-369. Mucch discussion and integration by the instructor, with a detailed prooofof the invariance theorem (theorem 4.3 [J95]).

February 26 (Tue), 2019 MD Shahriar Iqbal completes the presentation of: Anders L. Madsen and Finn Verner Jensen. "Lazy Propagation in Junction Trees." In: G.F. Cooper and S. Moral (eds.), Uncertainty in Artificial Intelligence: Proceedings of the Fourteenth Conference. San Francisco, CA: Morgan-Kaufmann, 1998, pp.362-369.
Detailed proof of theorem 4.6 [J95] with special care taken to explain the ("structural," on trees) induction step and the use of the invariance theorem.

February 28 (Thu), 2019 Sun (Yuxiang) Sun presents: Mark Bloemeke and Marco Valtorta. "A Hybrid Algorithm to Compute Marginal and Joint Beliefs in Bayesian Networks and Its Complexity." In: G.F. Cooper and S. Moral (eds.), Uncertainty in Artificial Intelligence: Proceedings of the Fourteenth Conference. San Francisco, CA: Morgan-Kaufmann, 1998, pp.16-23.
The following paper was added to the syllabus and to blackboard: Adnan Darwiche. "Inference by Conditioning." Chapter 8 in: Adnan Darwiche. Modeling and Reasoning with Bayesian Networks. Cambridge University Press, 2009

March 5 (Tue), 2019 Jianhai Su presents: Anders Madsen, Cory Butz, Jhonathan Oliveira, and Andre Dos Santos. "Simple Propagation with Arc Reversal in Bayesian Networks." Proceedings of the 9th International Conference on Probabilistic Graphical Models (PGM-18), PMLR-72 (Vaclav Kratochvil, Milan Studeny, editors), 2018, pp.260-271.

March 9 (Thu), 2019 The following paper was expected to be presented today, but will be presented on March 19, because Mr. Noah Geveke was absent.

  1. Finn Verner Jensen and Frank Jensen. "Optimal Junction Trees." Proceedings of the Tenth Conference on Uncertainty in Artificial Intelligence (UAI-1994), pp.360-366.
Discussion of some exercises from chapter 4 [J96].

March 19 (Tue), 2019 Noah Geveke presents:

  1. Finn Verner Jensen and Frank Jensen. "Optimal Junction Trees." Proceedings of the Tenth Conference on Uncertainty in Artificial Intelligence (UAI-1994), pp.360-366.
The following papers were assigned for presentation:
  1. Adnan Darwiche. "Inference by Conditioning." Chapter 8 in: Adnan Darwiche. Modeling and Reasoning with Bayesian Networks. Cambridge University Press, 2009. (To be presented by Sun (Yuxiang) Sun no earlier than Tuesday, March 26.)
  2. Marco Valtorta and Yimin Huang. "Identifiability in Causal Bayesian Networks: A Gentle Introduction." Cybernetics and Systems, 39, 4, 425-442, 2008. (To be presented by Rui XIn no earlier than Thursday, March 28.)
  3. Judea Pearl. "Causal Diagrams for Empirical Research." Biometrika, 82, 4, 669-710, 1995. (To be presented by Jianhai Su no earlier than Thurday, March 28.)
  4. Elizabeth S. Allman, John A. Rhodes, Elena Stanghellini, and Marco Valtorta. "Parameter Identifiability of Discrete Bayesian Networks with Hidden Variables." Journal of Causal Inference, 3, 2, pp.189-205, 2015 (referred to as [ARSV]). (To be presented by MD Shahriar Iqbal no earlier than Tuesday, April 2.)

March 21(Thu), 2019 Lecture on Conditional Independence and Markov Properties, based on ch.3 [L96]: conditional independence, the semigraphoid axioms, the graphoid axioms, the compositional graphoid axioms, Markov properites on undirected graphs: pairwise (P), local (L), and global (G); factorizazion (F).

March 26(Tue), 2019 Sun Sun presents the following paper:

  1. Adnan Darwiche. "Inference by Conditioning." Chapter 8 in: Adnan Darwiche. Modeling and Reasoning with Bayesian Networks. Cambridge University Press, 2009.
Proofs of the equivalence and non-equivalence of some Markov properties of undirected graphs (Markov networks).

March 28(Thu), 2019 Continuation of Ch.3 [L96], including proving the Clifford-Hammersley theorem and Moussouris's 1974 example that (G) does not imply (F) for non-positive distriutions.

April 2 (Tue), 2019 Completion of Ch.3 [L], including detailed explanation of the DAG (BN) Markov properties.

April 4 (Thu), 2019 Rui Xin begins presenting
Marco Valtorta and Yimin Huang. "Identifiability in Causal Bayesian Networks: A Gentle Introduction." Cybernetics and Systems, 39, 4, 425-442, 2008. Lecture on causal Bayesian networks, structural equations, and the excision semantics for intervention. The Coins and Bell example in detail. Very bried introduction to Pearl's graphical do-calculus of intervention.

April 9 (Tue), 2019 Guest lecture by Mr. Mohammad Ali Javidian: The back-door and front-door criteria.

April 11 (Thu), 2019 Intervention as a conditionalization in an augmented model. Detailed proof that the causal effect of X on Y is not identifiable in Fisher's genotype model of smoking and lung cancer. Rui Xin completes the presentation of
Marco Valtorta and Yimin Huang. "Identifiability in Causal Bayesian Networks: A Gentle Introduction." Cybernetics and Systems, 39, 4, 425-442, 2008.

April 16 (Tue), 2019 Jianhai Su presents: Judea Pearl. "Causal Diagrams for Empirical Research." Biometrika, 82, 4, 669-710, 1995. He also discusses surrogate experiments, using an example from a paper by Bareinboim and Pearl at UAI-2012. Students are asked to create an account in causalfusion.net; this is a beta version, which we are allowed to use by special permission of its developer, Dr. Elias Bareinboim.

April 18 (Thu), 2019 MD Shahriar Iqbal presents: Elizabeth S. Allman, John A. Rhodes, Elena Stanghellini, and Marco Valtorta. "Parameter Identifiability of Discrete Bayesian Networks with Hidden Variables." Journal of Causal Inference, 3, 2, pp.189-205, 2015 (referred to as [ARSV]).

April 23 (Tue), 2019 Project assigned (MENTOR reconstruction; latent variables in the MENTOR model; causal effects in the MENTOR model), due at the time of the final exam (May 7, 1600, in the classroom). Discussion of the project will take place of the final exam. Guest lecture by Mr. Mohammad Ali Javidian: The PC algorithm for structural learning of Bayesian networks.

April 25 (Thu), 2019 Guest lecture by Mr. Mohammad Ali Javidian: The PC algorithm for structural learning of Bayesian networks. Student course evaluation. End of course.