Spring Seminar Series - February 18, 2003
University of Minnesota
School of Statistics
College of Liberal Arts

From margin-based classification to ψ-learning

Xiaotong Shen
Department of Statistics
The Ohio State University

Tuesday, February 18, 2003
4:00 PM, 115 Ford Hall
Minneapolis, East Bank Campus
Social at 3:30 PM, 300 Ford Hall

Abstract

    
The concept of large margins plays an important role in analyzing learning methodologies such as Boosting, Neural Networks, and Support Vector Machines. In this talk, I will present a novel classification methodology called ψ-learning. While retaining the interpretation of large margins, ψ-learning delivers high performance in generalization, especially in nonseparable cases, as it is derived from a direct consideration of generalization errors. The nonconvex minimization involved in ψ-learning is solved via a global optimization method based on d.c. (differenced convex) programming. The theoretical foundation of ψ-learning will be discussed, in addition to illustrative examples.