CS Colloquium on Monday, May 17, 2004

Title: An Alternative Approach to Designing Clinical Trials:
       Budgeted Learning of Probablistic Classifiers

Speaker: Russ Greiner
         University of Alberta
         http://www.cs.ualberta.ca/~greiner/

Date: Monday, May 17

Time: 3PM

Place: MCS 135, 111 Cummington Street (please see
       http://cs-www.bu.edu/colloquium for directions)

Abstract:

Researchers often use clinical trials to collect the data needed to confirm
some hypothesis, or produce a classifier.  During this process, they have
to pay the cost of performing each test.  Many studies will run a
comprehensive battery of tests on each subject, for as many subjects as
their budget will allow.  We consider a more general model, where the
researcher can sequentially decide which single test to perform on which
specific individual; again subject to spending only the available funds.
Our goal here is to use these funds most effectively, to collect the data
that allows us to learn the most accurate classifier, We explore a number
of approaches -- both standard and novel -- and show that our novel
approaches (in particular "biased robin") are typically much more effective
that the standard "run every test" model.

See http://www.cs.ualberta.ca/~madani/BudgetedLearningPage.html

[This is joint work with Dan Lizotte and Omid Madani]

host: Simon Kasif