On the doubt about margin explanation of boosting

• Published in: Artificial intelligence Vol. 203; pp. 1 - 18
• Main Authors: Gao, Wei, Zhou, Zhi-Hua
• Format: Journal Article
• Language: English
• Published: Oxford Elsevier B.V 01.10.2013; Elsevier
• Subjects: Applied sciences; Boosting; Classification; Computer science; control theory; systems; Data processing. List processing. Character string processing; Ensemble methods; Exact sciences and technology; Margin theory; Memory organisation. Data processing; Software; Boosting; Ensemble methods; Margin theory; Classification; Data analysis; Automatic classification; Bounded error; Error bound; Generalization; Variance; Aggregate model; Bernstein polynomial; Voting; Supervised learning; Probability learning
• ISSN: 0004-3702; 1872-7921
• DOI: 10.1016/j.artint.2013.07.002

Margin theory provides one of the most popular explanations to the success of AdaBoost, where the central point lies in the recognition that margin is the key for characterizing the performance of AdaBoost. This theory has been very influential, e.g., it has been used to argue that AdaBoost usually does not overfit since it tends to enlarge the margin even after the training error reaches zero. Previously the minimum margin bound was established for AdaBoost, however, Breiman (1999) [9] pointed out that maximizing the minimum margin does not necessarily lead to a better generalization. Later, Reyzin and Schapire (2006) [37] emphasized that the margin distribution rather than minimum margin is crucial to the performance of AdaBoost. In this paper, we first present the kth margin bound and further study on its relationship to previous work such as the minimum margin bound and Emargin bound. Then, we improve the previous empirical Bernstein bounds (Audibert et al. 2009; Maurer and Pontil, 2009) [2,30], and based on such findings, we defend the margin-based explanation against Breimanʼs doubts by proving a new generalization error bound that considers exactly the same factors as Schapire et al. (1998) [39] but is sharper than Breimanʼs (1999) [9] minimum margin bound. By incorporating factors such as average margin and variance, we present a generalization error bound that is heavily related to the whole margin distribution. We also provide margin distribution bounds for generalization error of voting classifiers in finite VC-dimension space.