Advanced Computational Statistics

Autumn semester 2017

General information

Lecturer	Nicolai Meinshausen
Lectures	Thu 08-10 HG E 33.3 >>
Course catalogue data	>>

Course content

The course will look at various aspects of graphical models: introduction to graphical modelling, exact inference algorithms such as sum-product algorithm and applications to Hidden Markov Models, structure estimation and applications.

Announcements

September 1st, 2017:
Beginning of lecture: Thursday, September 21st.

Course materials

We will use various papers or book-chapters as sources, which are posted here as a password-protected link to polybox.


Introduction to the basic graph theoretical concepts
Markov Properties on graphs
Exact Inference algorithms and applications
Sum-product algorithm (belief propagation)
R file for Message-passing in simple HMM and Baum-Welch estimation (EM)
R help file for maze game
R maze game (forward-backward algorithm in filtering mode)
Notes on EM-algorithm
Maze game with noisy observations
Junction Tree and probabilistic expert systems
R example for junction-tree message-passing
Gaussian graphical models (without 1.3;1.4;1.5)
R example helper file for GGM example
R example for GGM completion of unobserved variables
Gaussian graphical model selection (glasso, clime etc.); sections 1.3 and 1.6
R example for graph selection with graphical Lasso and clime
R example for matrix completion
R example for covariance estimation via matrix completion for given graph (via dual problem)
Some additional voluntary reading: Structure Learning in Graphical Modelling Graphical Lasso paper ICA tutorial Matrix completion Nonparanormal modelling Score matching