This outline is still very preliminary, and is subject to change.  We'll put the latest updates to the syllabus here when they become available.  There are 29 lectures total.

Fundamentals of Bayesian Networks (directed graphical models): representation, semantics, learning and inference [11 Lectures]

  • Introduction to the class
  • Representation, semantics
  • MLE parameter learning
  • Structure learning for BNs - complete data
    • Constraint-based
    • Chow-Liu
    • Fixed-order
    • Structure search
  • Exact inference
    • Variable elimination
    • Junction trees
    • Context-specific independence

Mon., Sept. 08:

Wed., Sept. 10:

  • NO CLASS

Mon., Sept. 15:

Wed., Sept. 17:

Mon., Sept. 22:

Wed., Sept. 24:

Mon., Sept. 29:

  • NO CLASS

Wed., Oct. 1:

Mon., Oct. 6:

Wed., Oct. 8:

Mon., Oct. 13:

Wed., Oct. 15:

Mon., Oct. 20:

Wed., Oct. 22:

[Top]

Representation revisited [3 Lectures] 

  • Undirected models, Markov Random Fields
  • Factor graphs - unifying representation
  • Exponential family

Mon., Oct. 27:

Wed., Oct. 29:

[Top]

Inference revisited:approximate inference [3 Lectures]

  • Sampling
    • importance sampling
    • MCMC, Gibbs
  • Variational inference
  • Loopy belief propagation
    • Generalized belief propagation
    • Kikuchi

Mon., Nov. 3:

Wed., Nov. 5:

Mon., Nov. 10:

Wed., Nov. 12:

[Top]

Learning revisited [3 Lectures]

  • Parameter estimation in BNs with missing data
    • EM
    • Gradient descent
  • Structure learning for BNs - missing data
  • Learning undirected graphical models
    • Gradient algorithms
    • IPF for tabular MRFs
    • Structure Learning

Mon., Nov. 17:

Mon., Nov. 24:

[Top]

Special graphical models / Applications [6 Lectures]

  • Gaussian
    • Multivariate Gaussians
    • Gaussian graphical models
    • Inference in Gaussian models
    • Hybrid models
    • discrete and continuous variables
  • Hidden Markov Models
    • Representation
    • Inference
      • Forwards-Backwards
      • Viterbi
      • Baum-Welch
  • Kalman filter
    • representation
    • linearization
    • switching Kalman filter
    • assumed density filtering
  • DBNs
    • Representation
    • Inference
    • BK

Mon., Dec. 1:

[Top]

Advanced Topics [3 Lectures]

  • Causality
  • Relational probabilistic models
  • Template models

Mon., Dec. 3:

[Top]

Project Poster Session

3-6pm, Monday, Dec 1st

NSH Atrium

[Top]

Project Paper Due

Wednesday, Dec. 3rd by 3pm by email to the instructors list

[Top]

Final Exam

Out: Wednesday, Dec 3rd
Due: Wednesday, Dec 10th, Noon

[Top]