CONF More-Mayo97b/IDIAP Improved Pairwise Coupling Classification With Correcting Classifiers Moreira, Miguel Mayoraz, Eddy Nédellec, Claire Ed. Rouveirol, Céline Ed. EXTERNAL https://publications.idiap.ch/attachments/reports/1997/rr97-09.pdf PUBLIC https://publications.idiap.ch/index.php/publications/showcite/more-mayo97 Related documents Machine Learning: ECML-98 Lecture Notes in Artificial Intelligence 1398 160-171 1998 Springer April 1998 IDIAP-RR 97-09 The benefits obtained from the decomposition of a classification task involving several classes, into a set of smaller classification problems involving two classes only, usually called dichotomies, have been exposed in various occasions. Among the multiple ways of applying the referred decomposition, Pairwise Coupling is one of the best known. Its principle is to separate, in each binary subproblem, a pair of classes, ignoring the remaining ones, which causes the decomposition scheme to contain as much subproblems as the number of possible pairs of classes in the original task. Pairwise Coupling decomposition has so far been used in different applications. In this paper, various ways of recombining the outputs of all the classifiers solving the existing subproblems are explored, and an important handicap of its intrinsic nature is exposed, which consists in the use, for the classification, of impertinent information. A solution for this problem is suggested and it is shown how it can significantly improve the classification accuracy. In addition, a powerful decomposition scheme derived from the proposed correcting procedure is presented. REPORT More-Mayo97/IDIAP Improved Pairwise Coupling Classification With Correcting Classifiers Moreira, Miguel Mayoraz, Eddy EXTERNAL https://publications.idiap.ch/attachments/reports/1997/rr97-09.pdf PUBLIC Idiap-RR-09-1997 1997 IDIAP Proceedings of the 10th European Conference on Machine Learning, 1998 The benefits obtained from the decomposition of a classification task involving several classes, into a set of smaller classification problems involving two classes only, usually called dichotomies, have been exposed in various occasions. Among the multiple ways of applying the referred decomposition, Pairwise Coupling is one of the best known. Its principle is to separate, in each binary subproblem, a pair of classes, ignoring the remaining ones, which causes the decomposition scheme to contain as much subproblems as the number of possible pairs of classes in the original task. Pairwise Coupling decomposition has so far been used in different applications. In this paper, various ways of recombining the outputs of all the classifiers solving the existing subproblems are explored, and an important handicap of its intrinsic nature is exposed, which consists in the use, for the classification, of impertinent information. A solution for this problem is suggested and it is shown how it can significantly improve the classification accuracy. In addition, a powerful decomposition scheme derived from the proposed correcting procedure is presented.