(please see the main page for schedule information)
Semisuperised learning, i.e. statistical machine learning methods that use both labeled and unlabeled data, have been actively investigated in recent years. However, at the current stage, we still don't have a complete understanding of their effectiveness. This talk presents a closely related problem, which leads to novel and robust algorithms for semi-supervised learning. Specifically we consider the problem of learning what good classifiers (or predictors) are like from multiple learning tasks. We call this problem structural learning.
I will start with the structural learning problem in its simplest form, which covers the well-known Stein's effect, and move to the general Bayesian formulation. I will then present a more flexible and realistic formulation of the problem under the standard machine learning framework, in which theoretical analysis can be carried out. Some algorithms will then be proposed under this framework. We discuss numerical computational issues, and show that for a specific formulation, the problem can be solved by an iterative SVD procedure. Experiments will be given in the end to demonstrate the effectiveness of the proposed methods in the setting of semi-supervised learning.
Joint work with Rie Ando
Tong Zhang received a B.A. in mathematics and computer science from Cornell Universityin 1994 and a Ph.D. in computer Science from Stanford Universityin 1998. Since 1998, he has been with IBM Research, T.J. Watson Research Center, Yorktown Heights, New York, where he is now a research staff member in the Information and Knowledge Management department. His research interests include machine learning, numerical algorithms, and their applications.