An Uncertainty Framework for Classification

TL;DR

This paper introduces a unified uncertainty-based framework for classification, connecting probabilistic and possibilistic approaches, and demonstrating that SVMs are a special case of maximum-margin classifiers.

Contribution

It presents a generalized likelihood function based on uncertainty measures, unifying probabilistic and possibilistic classifiers, and reveals the relationship between SVMs and maximum-margin classifiers.

Findings

Probabilistic classifiers optimize cross-entropy.

Possibilistic classifiers maximize interclass margin.

Support vector machines are a subclass of maximum-margin classifiers.

Abstract

We define a generalized likelihood function based on uncertainty measures and show that maximizing such a likelihood function for different measures induces different types of classifiers. In the probabilistic framework, we obtain classifiers that optimize the cross-entropy function. In the possibilistic framework, we obtain classifiers that maximize the interclass margin. Furthermore, we show that the support vector machine is a sub-class of these maximum-margin classifiers.