Pairwise classifier combination using belief functions Export

@article{citeulike:1673185, abstract = {In the so-called pairwise approach to polychotomous classification, a multiclass problem is solved by combining classifiers trained to discriminate between each pair of classes. In this paper, this approach is revisited in the framework of the Dempster-Shafer theory of belief functions, a non-probabilistic framework for quantifying and manipulating partial knowledge. It is proposed to interpret the output of each pairwise classifiers by a conditional belief function. The problem of classifier combination then amounts to computing the non-conditional belief function which is the most consistent, according to some criterion, with the conditional belief functions provided by the classifiers. Experiments with various datasets demonstrate the good performances of this method as compared to previous approaches to the same problem.}, author = {Quost, Benjamin and Denoeux, Thierry and Masson, Marie-Helene}, citeulike-article-id = {1673185}, citeulike-linkout-0 = {http://dx.doi.org/10.1016/j.patrec.2006.11.002}, citeulike-linkout-1 = {http://www.sciencedirect.com/science/article/B6V15-4MK0HGM-1/2/7006b790735dcd3357409688a69ae0f9}, day = {1}, doi = {10.1016/j.patrec.2006.11.002}, journal = {Pattern Recognition Letters}, keywords = {classification, concept\_hierarchy, learning}, month = {April}, number = {5}, pages = {644--653}, posted-at = {2007-09-19 05:12:32}, priority = {2}, title = {Pairwise classifier combination using belief functions}, url = {http://dx.doi.org/10.1016/j.patrec.2006.11.002}, volume = {28}, year = {2007} }

TY - JOUR ID - citeulike:1673185 L3 - citeulike-article-id:1673185 N2 - In the so-called pairwise approach to polychotomous classification, a multiclass problem is solved by combining classifiers trained to discriminate between each pair of classes. In this paper, this approach is revisited in the framework of the Dempster-Shafer theory of belief functions, a non-probabilistic framework for quantifying and manipulating partial knowledge. It is proposed to interpret the output of each pairwise classifiers by a conditional belief function. The problem of classifier combination then amounts to computing the non-conditional belief function which is the most consistent, according to some criterion, with the conditional belief functions provided by the classifiers. Experiments with various datasets demonstrate the good performances of this method as compared to previous approaches to the same problem. IS - 5 JF - Pattern Recognition Letters EP - 653 TI - Pairwise classifier combination using belief functions VL - 28 SP - 644 KW - classification KW - concept_hierarchy KW - learning AU - Quost, Benjamin AU - Denoeux, Thierry AU - Masson, Marie-Helene PY - 2007/04/01/ UR - http://dx.doi.org/10.1016/j.patrec.2006.11.002 ER -