Evidential Multi-Label Classification Approach to Learning from Data with Imprecise Labels

Multi-label classification problems arise in many real-world applications. Classically, in order to construct a multi-label classifier, we assume the existence of a labeled training set, where each instance is associated with a set of labels, and the task is to output a label set for each unseen instance. However, it is not always possible to have perfectly labeled data. In many problems, there is no ground truth for assigning unambiguously a label set to each instance, and several experts have to be consulted. Due to conflicts and lack of knowledge, labels might be wrongly assigned to some instances. This paper describes an evidence formalism suitable to study multi-label classification problems where the training datasets are imperfectly labelled. Several applications demonstrate the efficiency of our apporach.

[1]  Jiebo Luo,et al.  Learning multi-label scene classification , 2004, Pattern Recognit..

[2]  Volker Roth,et al.  Kernel methods for regression and classification , 2001 .

[3]  Philippe Smets,et al.  The Combination of Evidence in the Transferable Belief Model , 1990, IEEE Trans. Pattern Anal. Mach. Intell..

[4]  Zhi-Hua Zhou,et al.  ML-KNN: A lazy learning approach to multi-label learning , 2007, Pattern Recognit..

[5]  Min-Ling Zhang,et al.  Ml-rbf: RBF Neural Networks for Multi-Label Learning , 2009, Neural Processing Letters.

[6]  Eyke Hüllermeier,et al.  Multilabel classification via calibrated label ranking , 2008, Machine Learning.

[7]  Thierry Denoeux,et al.  Representing uncertainty on set-valued variables using belief functions , 2010, Artif. Intell..

[8]  Jeff Z. Pan,et al.  An Argument-Based Approach to Using Multiple Ontologies , 2009, SUM.

[9]  Thierry Denoeux.  A k -Nearest Neighbor Classification Rule Based on Dempster-Shafer Theory , 2008, Classic Works of the Dempster-Shafer Theory of Belief Functions.

[10]  Grigorios Tsoumakas,et al.  Random k -Labelsets: An Ensemble Method for Multilabel Classification , 2007, ECML.

[11]  Grigorios Tsoumakas,et al.  Multi-Label Classification: An Overview , 2007, Int. J. Data Warehous. Min..

[12]  Philippe Smets,et al.  Classification Using Belief Functions: Relationship Between Case-Based and Model-Based Approaches , 2006, IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics).

[13]  Thierry Denoeux,et al.  Multi-label classification algorithm derived from K-nearest neighbor rule with label dependencies , 2008, 2008 16th European Signal Processing Conference.

[14]  Joost N. Kok.  Machine Learning: ECML 2007, 18th European Conference on Machine Learning, Warsaw, Poland, September 17-21, 2007, Proceedings , 2007, ECML.

[15]  Thierry Denoeux,et al.  An Evidence-Theoretic k-Nearest Neighbor Rule for Multi-label Classification , 2009, SUM.