Date
|
Summary |
06/01/2014
|
Overview of the course
|
09/01/2014--20/01/2014
|
Representation: graphical models (Chapter 3 of KF)
-
Directed graphical models: overview, representation of probability distribution and conditional independence statements.
|
16/01/2014
|
Quiz
Syllabus: Chapter 2 and material covered up to 13/01/2014.
|
23/01/2014--30/01/2014
|
Undirected graphical models (Chapter 5 of KF)
- Representation: potentials, conditional independence and graph separability, factorization.
- Constructing undirected models from distributions
- Relationship between directed and undirected models.
- Common undirected graphical models: Factor models, Ising and Potts model, Gibbs distribution, log-linear models, CRFs.
- Feature-based potentials for flexible deployment in many applications.
- Application in vision and text mining.
|
03/02/2014 --- 13/02/2014
|
Exact Inference
-
Overview (Chapter 9 of KF)
-
Variable elimination (Chapter 9 of KF)
- Junction trees and sum product message passing (Chapters 10 of KF)
- Creating junction trees, Junction trees: triangulation and decomposability (Chapters 10 of KF)
|
|
Midsems week
|
24/02/2014 --- 27/02/2014
|
Combinatorial algorithms for MAP inference. (Chapter 13 in KF)
- Fast Approximate Energy Minimization via Graph Cuts,
Yuri Boykov, Olga Veksler and Ramin Zabih, IEEE Transactions on PAMI, vol. 23, no. 11, pp. 1222-1239. local copy
- What energy functions can be minimized via graph cuts. local copy
|
03/03/2014--10/03/2014
|
Inference via sampling. (Chapter 12 in KF)
- Forward sampling, rejection sampling, and likelihood weighted sampling in Bayesian networks
- Importance sampling and relationship to likelihood weighted sampling
- MCMC and Gibbs sampling in general graphical models
|
13/03/2014---27/04/2014
|
Learning graphical model parameters (probabilistic methods)
- Parameter estimation in Bayesian networks (completely observed variable set): MLE (Chapter 16, 17.1--17.4.3 )
- Local CPDs: exponential familiy (Chapter 8 from KF)
- Parameter estimation in undirected graphs (Chapter 20.1--20.2.3)
- Learning conditional graphical models (CRFs), conditional likelihood training. (Chapter 20.3.1--20.3.2)
- Learning with partially observed data, training a HMM
|
27/03/2014--03/04/2014
|
Structured learning.
|
07/04/2014---10/04/2014
|
Learning graphical model structure
|
-->
14/04/2014 -- 14/04/2014
|
Approximate inference based on loopy belief propagation
- Chapter 11.3.1--11.3.5 of KF
|