Consistent Multicategory Support Vector Machines

• Activity Number: 355
• Type: Invited
• Date/Time: Wednesday, August 1, 2007 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statisticians in Defense and National Security
• Author(s): Yufeng Liu and Ming Yuan*+
• Companies: The University of North Carolina at Chapel Hill and Georgia Institute of Technology
• Keywords: classification ; consistency ; margin ; multicategory

Abstract

The Support Vector Machine (SVM) has become one of the most popular machine learning techniques in recent years. The success of SVM is mostly due to its elegant margin concept and theory in binary classification. Generalization to the multicategory setting, however, is not trivial. There are a number of different multicategory extensions of SVM in the literature. In this talk, we review several commonly used extensions and explore Fisher consistency of these extensions. A new class of multicategory SVMs are proposed. We develop efficient numerical algorithms to implement the proposed methods. Theoretical properties are explored as well.