On the relationship between majority vote accuracy and dependency in multiple classifier systems


We define two simple probability models of dependency between classifiers: a multiplicative form and an additive form. This paper explores the relationship between the majority vote accuracy and dependency for the three classifiers. We show that the majority votes with negatively dependent classifiers can offer an improvement over independent classifiers… (More)
DOI: 10.1016/S0167-8655(02)00260-X


4 Figures and Tables