AUEB STATS SEMINARS 16-20/4/2018: A short Course on GRAPHICAL MODELS by Prof. Helene Massam

Facebook event: https://www.facebook.com/events/2013639165542646/
Free Registration: https://goo.gl/forms/WBfrf92k7FHVBO3v2 [~20 positions - deadline 30/3/2018]

AUEB STATISTICS SEMINAR SERIES APRIL 2018

SHORT COURSE

Helene Massam
Professor, Department of Mathematics and Statistics
York University, Canada

A short Course on GRAPHICAL MODELS

• Lecture 1 [Monday 16 April 2018 12.00-15.00] Introduction and Basic Notions.
  
• Lecture 2 [Tuesday 17 April 2018 9.00-12.00] Graphical Gaussian models.
  
• Lecture 3 [Wednesday 18 April 2018 9.00-12.00] Discrete graphical and hierarchical models.
  
• Lecture 4 [Thursday 19 April 2018 9.00-12.00] Model selection and learning for graphical Gaussian models.
  
• Lecture 5 [Friday 20 April 2018 9.00-12.00] Model selection and learning for discrete graphical  models.
  

All lectures will be placed in Room 802, 8th floor of the Postgraduate Building of Athens University of Economics and Business (Evelpidon & Lefkados).  

• The course is financed by the M.Sc. in Statistics of Athens University of Economics and Business.
• A limited number of positions (~20) will be available for other participants (outside the Full time program of the  M.Sc. of Statistics) with preference given to Ph.D. students and M.Sc. Students and graduates (with this order).
• Free Registration is available for a limited number until 30/3/2018 or earlier if the positions are covered at https://goo.gl/forms/WBfrf92k7FHVBO3v2
• Your position will be secured only after official notification by the Postgraduate office of Statistics of AUEB.  
• Certificate of attendance will be provided (electronically) to all participants attending at least 4 out of 5 lectures.
 

Detailed Structure of the Course

• Lecture 1: Introduction and Basic notions
  
  
  
  • (a) graph theory,
    
  • (b) conditional independence and Markov properties
    
  • (c) exponential families
    
  • (d) Elementary examples of graphical models.
    
  
  
  
• Lecture 2: Graphical Gaussian models:
  
  
  
  • a. Undirected decomposable and non decomposable models, directed models.
    
  • b. The Wishart distribution.
    
  • c. Parameter estimation through maximum likelihood and Bayesian methods.
    
  • d. Sampling from conjugate priors.
    
  
  
  
• Lecture 3: Discrete graphical and hierarchical models
  
  
  
  • a. Undirected decomposable and non decomposable models, directed models.
    
  • b. The Dirichlet distribution.
    
  • c. Parameter estimation through maximum likelihood and Bayesian methods.
    
  • d. Sampling from conjugate priors.
    
  
  
  
• Lecture 4: Model selection and learning for graphical Gaussian models.
  
  
  
  • a. Basic model selection methodology from the frequentist point of view: the G-Lasso.
    
  • b. Basic model selection methodology from the Bayesian point of view: The Bayes factor; computation of the normalizing constant in Gaussian models, travelling through the set of graphs.
    
  • c. A review of sampling methods for sampling from the G-Wishart.  
    
  
  
  
• Lecture  5:  Model selection and learning  for discrete graphical  models.
  
  
  
  • a. Computation of the normalizing constant in discrete graphical models
    
  • b. A review of sampling methods for sampling from the G-Wishart and the generalized hyper- Dirichlet.
    
  • c. Moving away from Bayes factors for model selection: a survey of recent methods for model
    
  
  
  

References
• Graphical models by Steffen Lauritzen, Oxford Science publications, 1996.
• Probabilistic graphical models by Kohler and Friedman, Springer, 2009.



These references and the ones listed below are only for guidance. All lectures will be based on my lecture notes.

Σας ενημερώνω ότι ήδη έχουμε δεχτεί διπλάσιο αριθμό αιτήσεων από αυτό που μπορούμε να φιλοξενήσουμε. Όσοι ήδη έχετε κάνει αίτηση, θα λάβετε μήνυμα αποδοχής ή όχι μέχρι τις 5/4/2018.