Comparison of two topological approaches for dealing with noisy labeling

Fabien A Rico (1) , Fabrice Muhlenbach (2) , Djamel Abdelkader Zighed (1) , Stéphane A Lallich (1)

1 ERIC - Equipe de Recherche en Ingénierie des Connaissances
2 LHC - Laboratoire Hubert Curien

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.

Mots clés: Identification of mislabeled instance; relaxation; cut edges weighted; topological learning; geometrical graphs; separability index; machine learning

Domaines: Apprentissage [cs.LG]

Fabien A Rico, Fabrice Muhlenbach, Djamel Abdelkader Zighed, Stéphane A Lallich. Comparison of two topological approaches for dealing with noisy labeling. Neurocomputing, 2015, 160, pp.3 - 17. ⟨10.1016/j.neucom.2014.10.087⟩. ⟨hal-01524431⟩