PhD Qualifying Examination


"Learning Latent Variable Models: A Survey"

Mr. Kin Man Poon


Abstract:

Latent variable models have been employed for a few decades in some
science fields, such as biometrics, psychometrics, and econometrics. The
usual approach in these areas is to specify a hypothesis with a latent
variable model and test whether this model fits the data. It assumes
some theoretical foundation to specify the hypothesis and the model has
to be modified manually to improve the model fit. This manual approach
is different from the approach commonly used in the field of machine
learning, which tries to discover knowledge automatically from data
without much prior knowledge about the domain.

This papers surveys some types of latent variable models usually used in
the social science field. It also looks at some recent attempts to learn
graphical models with hidden variables in the machine learning field. It
aims to explore research directions that combine both fields of research
so that some interesting types of latent variable models can be learned
automatically from data.


Date:     		Monday, 28 August 2006

Time:                   2:30p.m.-4:30p.m.

Venue:                  Room 3311
			lifts 17-18

Committee Members:      Dr. Nevin Zhang (Supervisor)
			Dr. Mordecai Golin (Chairperson)
			Dr. Brian Mak
			Dr. Dit Yan Yeung


**** ALL are Welcome ****