Spring Seminar Series - February 3, 2003
University of Minnesota
School of Statistics
College of Liberal Arts
Complexity regularization via localized random penalties
Marten Wegkamp
Department of Statistics
Yale University
Monday, February 3, 2003
4:00 PM, 130
Ford Hall
Minneapolis, East Bank Campus
Social at 3:30 PM, 300
Ford Hall
Abstract
In this joint work with Gabor Lugosi (Pompeu Fabra University) model selection
via penalized empirical loss minimization in nonparametric classification
problems is studied. Data-dependent penalties are constructed, which are
based on estimates of the complexity of a small subclass of each model class,
containing only those functions which have small empirical loss. The penalties
are novel since the penalties considered in the literature are typically
based on the entire model class. Oracle inequalities using these penalties
are established, and the advantage of the new penalties over the penalties
based on the complexity of the whole model class is demonstrated.