Cambridge/Gatsby Machine Learning Workshop
Sampling
Thursday 8th January 2004
DRAFT SCHEDULE
Format: Each speaker is asked to lead a discussion, starting with
a 20-40 minute talk highlighting issues for discussion.
To ensure that our discussions don't run out of juice,
everyone who attends is asked to bring one issue or question for discussion.
Recommended reading links are included with some of the talks
below. In addition, everyone attending should be familiar with
the following background material:
importance sampling,
rejection sampling,
Metropolis method,
Gibbs sampling;
familiarity with
slice sampling
and Hamiltonian Monte Carlo may also prove helpful.
All these topics are reviewed in
chapters 29 and 30 of DJCM's book.
Time
| Discussion leader
| Topic
| Recommended reading |
10.00
|
Andrew Blake
|
Particle filters
|
Reading
|
11.00
| Coffee
|
11.30 |
Iain Murray |
Contrastive divergence
| Reading |
12.30
| Discussion
|
12.45
| Lunch
|
|
| 2.00
|
Yuan Qi
| Hessian based MCMC
| Reading | 3.00
|
Zoubin Ghahramani |
Embedded HMMs, an idea
for 'particle smoothing'.
| Radford Neal paper
| 4.00
| Tea
| 4.30
|
David MacKay |
Exact sampling |
Reading |
5.30
| Discussion
|
|