%0 Journal Article
%T Comparison of two topological approaches for dealing with noisy labeling
%+ Equipe de Recherche en Ingénierie des Connaissances (ERIC)
%+ Laboratoire Hubert Curien [Saint Etienne] (LHC)
%A Rico, Fabien, A
%A Muhlenbach, Fabrice
%A Zighed, Djamel, A
%A Lallich, Stéphane, A
%< avec comité de lecture
%@ 0925-2312
%J Neurocomputing
%I Elsevier
%V 160
%P 3 - 17
%8 2015
%D 2015
%R 10.1016/j.neucom.2014.10.087
%K Identification of mislabeled instance
%K relaxation
%K cut edges weighted
%K topological learning
%K geometrical graphs
%K separability index
%K machine learning
%Z Computer Science [cs]/Machine Learning [cs.LG]Journal articles
%X This paper focuses on the detection of likely mislabeled instances in a learning dataset. In order to detect potentially mislabeled samples, two solutions are considered which are both based on the same framework of topological graphs. The first is a statistical approach based on Cut Edges Weighted statistics (CEW) in the neighborhood graph. The second solution is a Relaxation Technique (RT) that optimizes a local criterion in the neighborhood graph. The evaluations by ROC curves show good results since almost 90% of the mislabeled instances are retrieved for a cost of less than 20% of false positive. The removal of samples detected as mislabeled by our approaches generally leads to an improvement of the performances of classical machine learning algorithms.
%G English
%2 https://hal.archives-ouvertes.fr/hal-01524431/document
%2 https://hal.archives-ouvertes.fr/hal-01524431/file/Neurocomputing__SL_FM_FR_DAZ_V5.pdf
%L hal-01524431
%U https://hal.archives-ouvertes.fr/hal-01524431
%~ CNRS
%~ ERIC
%~ UNIV-LYON2
%~ UNIV-ST-ETIENNE
%~ IOGS
%~ PARISTECH
%~ IOGS-SACLAY
%~ UNIV-PARIS-SACLAY