An Uncertainty Framework for Classification
This work provides a unified uncertainty framework for classification, which is incremental as it generalizes existing methods like cross-entropy and support vector machines.
The paper tackles the problem of classification by proposing a generalized likelihood function based on uncertainty measures, showing that maximizing it leads to different classifiers: probabilistic ones optimize cross-entropy, while possibilistic ones maximize interclass margin, with support vector machines as a sub-class.
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.