MCMC Journal Club

(Organised by Carl Edward Rasmussen, Cyril Goutte, Henrik Oeyelund & Rod Murray-Smith)

Every  Wednesday 13:00-15:00 in Room 205 in Building 305, starting Sept. 9th 1998.
 

Timetable

We will begin with a few introductory papers, so no deep prior insight will be required. At each session a particular paper will be discussed. The discussion will be lead by one person chosen at random at the beginning of the meeting. We think that this scheme will lead to the best discussions and most thorough understanding, through better preparation. We plan that participation in  the journal-club can be credited in terms of "points"  towards the PhD requirements by being included in the students Studievejledning.

Everybody on the mailing list will be provided with photocopies of the notes in advance (these will be left outside room 206 in building 305).
 
 
Sept. 9.   S. Brooks,"Markov Chain Monte Carlo  and its application", The Statistician, 1998, 47, Part1, pp 69-100. 
(Lead discussion article will be presented by Carl Rasmussen ) 
David Mackay, Introduction to MCMC 
Chib & Greenberg, Understanding the Metropolis-Hastings Algorithm 
Kass, R.E., B.P. Carlin, A. Gelman and R.M. Neal (February, 1997) Markov Chain Monte Carlo in Practice: A Roundtable Discussion. Department of Statistics, Carnegie Mellon University. Available as Compressed Postscript. and pdf
Sept. 16. Bo talking about Markov Chains
Sept. 23. Mackay 1-17
Sept. 30. Rest of MacKay paper, Implementation of linear model, 
BUGS,
W. Gilks, Full conditional distributions, 
Oct 7. Convergence diagnostics Go for the Cowles & Carlin 96 paper
Oct 14. Holiday
Oct 21 Lars Kai Hansen, Mean Field approaches. Get the m-files for the talk. File1 File2.
Oct 28. Harpa/Henrik, Simple Mixture models,
Jean Diebolt, Chr. Robert, Estimation of Finite Mixture Distributions through Bayesian Sampling, 
J.R Statist. Soc B 1994, No 2, p363-375.
Nov  4. Pedro, Variable dimension MCMC, 
Richardson & Green,On Bayesian analysis of mixtures  with an unknown number of components (with discussion)....
Nov 11 Cyril, Exact Sampling
Propp and Wilson. ``Exact sampling with coupled Markov chains and applications to statistical mechanics.'' Random Structures and Algorithms, 9(1&2):223--252, 1996

There is also a nice Coupling from the past Java demo
m-files for the talk: mcsample.m, mcexact.m (3 state Markov chain); hardcore.m (Hardcore model), hardcorev.m (same with runtime output + figures)

Nov 18. Carl, Infinite Mixture models
Nov 25. Rune, Texture segmentation
....
Jan, MCMC modelling of SDEs. (date depends on workload!)
If you haven't heard about MCMC, links can be found here:

Participants

Carl Edward Rasmussen, Cyril Goutte, Henrik Oeyelung, Roderick Murray-Smith, Poul Thyregaard, Jan Larsen, Allan Nielsen, Niels Herman Hansen, Pedro H. Soerensen,  Rune Fisker, Bo Friis Nielsen, Annette Kjaer Ersboell, Lars Kai Hansen, Finn Nielsen, Mikel Baadsgaard, Niels Petersen,  Harpa Jonsdottir, Helle Holst, Ulrik Kjems.

If you want to join, just mail us.

General Background

During the coming fall and spring terms, we're planning to make an effort
to learn about Markov Chain Monte Carlo (MCMC) methods. The idea has been
motivated by the fact that 1) separate efforts are being made in nearly
every section of IMM to exploit MCMC ideas and 2) MCMC methods are
currently a rapidly expanding area, both theoretically and in terms of
applications.

We're planning to have a journal-club in the fall and in the spring a more
formalized course with invited speakers (leading MCMC experts). We are
hoping that both PhD student and VIP's will participate, and that besides
learning about MCMC methods the course will stimulate cooperation between
the sections of IMM.

In the spring, a more formal course will be arranged with speakers both
from IMM and with invited speakers. Hands-on projects inspired by the
ongoing work in the sections will also be arranged. This course will be
lead by Poul Thyregod and Henrik Madsen (both IMM).

Please send email to Rod Murray-Smith (rod@imm.dtu.dk) if you might be
interested in participating, and if you have further suggestions.