An Empirical Study of the Convergence of RegionBoost
RegionBoost is one of the classical examples of Boosting with dynamic weighting schemes. Apart from its demonstrated superior performance on a variety of classification problems, relatively little effort has been devoted to the detailed analysis of its convergence behavior. This paper presents some results from a preliminary attempt towards understanding the practical convergence behavior of RegionBoost. It is shown that, in some situations, the training error of RegionBoost may not be able to converge consistently as its counterpart AdaBoost and a deep understanding of this phenomenon may greatly contribute to the improvement of RegionBoost.
KeywordsBoosting RegionBoost Convergence kNN Decision Stumps
Unable to display preview. Download preview PDF.
- 1.Dietterich, T.G.: Machine Learning Research: Four Current Directions. AI Magazine 18(4), 97–136 (1997)Google Scholar
- 8.Schapire, R.E.: The Strength of Weak Learnability. Machine Learning 5(2), 197–227 (1990)Google Scholar
- 9.Maclin, R.: Boosting Classifiers Regionally. In: The 15th National Conference on Artificial Intelligence, Madison, WI, pp. 700–705 (1998)Google Scholar
- 11.Moerland, P., Mayoraz, E.: DynaBoost: Combining Boosted Hypotheses in a Dynamic Way. In: IDIAP-RR, Switzerland (1999)Google Scholar
- 13.Jin, R., Liu, Y., Si, L., Carbonell, J., Hauptmann, A.G.: A New Boosting Algorithm Using Input-Dependent Regularizer. In: The 20th International Conference on Machine Learning, Washington, DC (2003)Google Scholar
- 15.Asuncion, A., Newman, D.J.: UCI Machine Learning Repository. University of California, School of Information and Computer Science (2007), http://www.ics.uci.edu/~mlearn/MLRepository.html