Skip to content
conftrace_
2004 JMLR JMLR 2004

Probability Estimates for Multi-class Classification by Pairwise Coupling

Abstract

Pairwise coupling is a popular multi-class classification method that combines all comparisons for each pair of classes. This paper presents two approaches for obtaining class probabilities. Both methods can be reduced to linear systems and are easy to implement. We show conceptually and experimentally that the proposed approaches are more stable than the two existing popular methods: voting and the method by Hastie and Tibshirani (1998). [abs] [ pdf ][ ps.gz ][ ps ]

📈 Trend Setter - Support Vector Machine
🧭 Keyword Pioneer - linear system
🐣 Hot Topic Early Bird - support vector machine

Authors

Ting-Fan Wu, Chih-Jen Lin, Ruby C. Weng

Topics

Machine Learning > Core Methods > Classification Machine Learning > Core Methods > Support Vector Machine

Keywords

multi-class classification support vector machine linear system pairwise coupling class probability discriminant classification
Download PDF

More from JMLR 2004

All JMLR 2004 papers →
Some Dichotomy Theorems for Neural Learning Problems
Image Categorization by Learning and Reasoning with Regions
Boosting as a Regularized Path to a Maximum Margin Classifier
On Robustness Properties of Convex Risk Minimization Methods for Pattern Recognition
Rational Kernels: Theory and Algorithms
Reinforcement Learning with Factored States and Actions