Professor:
Luis Enrique Sucar
esucar (AT) inaoep.mx

L. E. Sucar, Probabilistic Graphical Models: Principles and Applications, Springer 2015

Program

Class Sessions

• Read paper on Reasoning under Uncertainty (in References, 3 parts)
• Read paper on Interpretations of Probability (in References, Philosophical aspects)
  
• Read Chapter 1 - Textbook
  

• Read Chapter 2 - Textbook
• Solve the following problems from  Chapter 2:  4,5,8,9
  

• Read Chapter 3 - Textbook
• Solve the following problems from  Chapter 3: 1,7,8,9
  

• Read Chapter 4 - Textbook
• Solve the following problems from  Chapter 4: 1, 5, 9 (Problem 9 should be handin next class)
  

• Read Chapter 5 - Textbook
• Solve the following problems from  Chapter 5: 1, 3, 4, 5
  

• Read Chapter 6 - Textbook
• Solve the following problems from  Chapter 6: 1, 4, 5

• Read Chapter 7 - Textbook
• Solve the following problems from  Chapter 7: 1, 3, 5

• Final Project Proposal: Handin a one page summary of your final project proposal, including: (i) problem description, (ii) objective, (iii) PGM techniques to be implemented, (iv) expected results.
• Download the demo version of "Hugin" (software for Bayesian nets). Using this software: (a) Construct using the graphical interface a BN for the example in Fig. 7.1 of the text book, defining the values for each variable and its CPTs. (b) Perform several inferences give different evidences and obtaing the posterior probabilities of all the variables. (c) Use the learning tool to obtain the structure and parameters for one of the dataset of problem 9, chapter 4. Handin a written report with the graphs and results of each part.
  

• Read Chapter 8 - Textbook
• Solve the following problems from  Chapter 8: 1, 2, 7
  

• It covers from Chapter 1 to Chapter 8     
  
• Two parts: (i) Theoretical: questions about the main concepts covered in the course, (ii) Practical: problems related to the techniques analized in the course (similar to the homework problems)
•  Notes: you can use one letter size page of notes during the exam; any other material is not allowed

• Final project: report and presentation
  
• Report: scientific paper (Format Lecture Notes from Springer) maximum 10 pages (without considering appendix) with the following (send via e-mail and handin printout):
  
  • Abstract
  • Introduction
  • Related work
    
  • Methodology
    
  • Experiments and results
    
  • Conclusions and future work
    
  • References (standar format)
  • Appendix (detailed results, code, ...)
• Presentation: 15 Minutes presentation including:
  
• Problem description
• Related work
  
• Methodology
• Results
• Conclusions

Philosofical aspects

• Interpretations
• Dutch Book arguments
  

Reasoning under uncertainty

• Part 1
• Part 2
• Part 3

• Introduction to Graphical Models (K. Murphy) 
  

• PageRank in Google
• HMMs: Tutorial by Rabiner
  

• Introducción a Redes Bayesianas-In Spanish (capítulo de libro: Sierra (Ed.), Aprendizaje Automático, Pearson, 2006)
• Loopy Propagation
  

• Application in Power Plants
  
• POMDP example
  

• Online proceedings
  

• Weka
• Hugin
• Elvira
• BNs in MatLab
• MDPs in MatLab
• ProBT
• deal - Learning BNs
• GOBNILP - C program for learning BN
  

Software/Data:

• Software for Graphical Models
  
• Tools for learning AI (AISpace - UBC)
• UC Irvine Machine Learning Repository
  

• UAI association
  

• UAI 2001
• UAI 2002
• UAI 2003
• UAI 2004
• UAI 2005
• UAI 2006
• UAI 2007
• UAI 2008
• UAI 2009
• UAI 2010
• UAI 2011
• UAI 2012
• UAI 2013
• UAI 2014
• UAI 2015
• UAI 2016
• UAI 2017
• UAI 2018
• UAI 2019
• UAI 2020
  

• PGM 2002
• PGM 2004
  
• PGM 2006
• PGM 2008
• PGM 2010
• PGM 2012
• PGM 2014
• PGM 2016
• PGM 2018
• PGM 2020
  

• FLAIRS 2008
• FLAIRS 2009
• FLAIRS 2010
• FLAIRS 2011
• FLAIRS 2012
• FLAIRS 2013
• FLAIRS 2014
• FLAIRS 2015
• FLAIRS 2016
• FLAIRS 2017
• FLAIRS 2018
• FLAIRS 2019
• FLAIRS 2020
  

• General information, tools, ...

• Develop a more efficient way to learn semi-naive Bayesian classifiers
  
• Develop a SLAM (simultaneous localization an mapping) based on graphical models
  
• Object tracking with Kalman / Particle Filters
  
• Learning dynamic Bayesian networks
  
• Implement "testing Bayesian networks" (more expressive, similar to NNs)
  
• Implement inference (prediction, counterfactuals) for Bayesian causal networks
  
• Learn temporal / dynamic causal networks
  

• Imágenes multiresolución (CAM)
  
• Gramática visual (redes bayesianas)
• Clasificación de texto (clasificador bayesiano)
• Evasión de obstáculos para robots (MDPs)
• Detección de fallas (redes bayesianas)
• Etiquetado de imágenes (clasificador bayesiano)
• Clasificación de huellas (clasificadores bayesianos)
• Ordenamiento de imágenes (CAM)
• Clasificación de texto (CAM)
• Clasificación de texto (redes bayesianas)
• Clasificadores jerárquicos (redes bayesianas)
• Búsqueda de objetos con un manipulador robótico (MDPs)
• Aprendizaje de redes bayesianas temporales (redes bayesianas)
• Clasificación de mutaciones para HIV (clasificadores bayesianos)
• Marcas de agua robustas (HMMs)
• Clasificación de objetos en imágenes (redes bayesianas)
• Clasificación jerárquica de imágenes (clasificadores y redes bayesianas)
• Clasificación de documentos (clasificadores bayesianos)
• Clasificadores bayesianos en cadena (TAN)
• Reconocimiento de ademanes (HMMs)
• Clasificadores bayesianos para la detección de engaño (clasificadores)
• Reconocimiento de ademanes usando HMMs
• Aprendizaje por refuerzo en juegos (MDPs)
• Seguimiento de personas usando el filtro de Kalman (redes bayesianas dinámicas)