Learning with two imperfect teachers using mathematical theory of evidence

Abstract

The major contribution of this work is the introduction of the concept of learning in an uncertain environment wherein the observed samples are classified by two independent classifiers. In contrast to other prevalent schemes, the authors' learning process incorporates the decisions of both the classifiers in a uniform manner. The algorithms handle two different types of uncertainty in the classification process: uncertainty in classification by a classifier and the uncertainty in the values of parameters of the classifier. For a human classifier (called the teacher), the uncertain parameter is his degree of imperfectness. The results of simulation on Iris data show that the learning of the uncertain parameters in the classifier can be only as good as the decisions given by the classifiers.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>