A Refined Margin Analysis for Boosting Algorithms via Equilibrium Margin
A Refined Margin Analysis for Boosting Algorithms via Equilibrium Margin
复制标题
通过均衡保证金提升算法的精细保证金分析
DOI:
10.5555/1953048.2021058
复制
发表时间:
2011-02
影响因子:
6
通讯作者:
Feng, Jufu
中科院分区:
文献类型:
--
作者:
Wang, Liwei;Sugiyama, Masashi;Jing, Zhaoxiang;Yang, Cheng;Zhou, Zhi-Hua;Feng, Jufu
Much attention has been paid to the theoretical explanation of the empirical success of AdaBoost. The most influential work is the margin theory, which is essentially an upper bound for the generalization error of any voting classifier in terms of the margin distribution over the training data. However, important questions were raised about the margin explanation. Breiman (1999) proved a bound in terms of the minimum margin, which is sharper than the margin distribution bound. He argued that the minimum margin would be better in predicting the generalization error. Grove and Schuurmans (1998) developed an algorithm called LP-AdaBoost which maximizes the minimum margin while keeping all other factors the same as AdaBoost. In experiments however, LP-AdaBoost usually performs worse than AdaBoost, putting the margin explanation into serious doubt. In this paper, we make a refined analysis of the margin theory. We prove a bound in terms of a new margin measure called the Equilibrium margin (Emargin). The Emargin bound is uniformly sharper than Breiman's minimum margin bound. Thus our result suggests that the minimum margin may be not crucial for the generalization error. We also show that a large Emargin and a small empirical error at Emargin imply a smaller bound of the generalization error. Experimental results on benchmark data sets demonstrate that AdaBoost usually has a larger Emargin and a smaller test error than LP-AdaBoost, which agrees well with our theory.
登录
查看更多内容
影响因子:
1.6
作者:
L. Devroye
通讯作者:
L. Devroye
DOI:
10.5555/1390681.1390687
发表时间:
2008-06
期刊:
J. Mach. Learn. Res.
影响因子:
--
作者:
David Mease;A. Wyner
通讯作者:
David Mease;A. Wyner
DOI:
10.5555/1005332.1044712
发表时间:
2004-12
期刊:
J. Mach. Learn. Res.
影响因子:
--
作者:
C. Rudin;I. Daubechies;R. Schapire
通讯作者:
C. Rudin;I. Daubechies;R. Schapire
DOI:
10.1198/016214505000000907
发表时间:
2006-03-01
影响因子:
3.7
作者:
Bartlett, PL;Jordan, MI;McAuliffe, JD
通讯作者:
McAuliffe, JD
影响因子:
7.5
作者:
Dietterich, TG
通讯作者:
Dietterich, TG