Boosting in the Presence of Noise

Journal of Computer and System Sciences | , Vol 71(3): pp. 266-290

View Publication | View Publication

Boosting algorithms are procedures that “boost” low-accuracy weak learning algorithms to achieve arbitrarily high accuracy. Over the past decade boosting has been widely used in practice and has become a major research topic in computational learning theory. In this paper we study boosting in the presence of random classification noise, giving both positive and negative results.

We show that a modified version of a boosting algorithm due to Mansour and McAllester (J. Comput. System Sci. 64(1) (2002) 103) can achieve accuracy arbitrarily close to the noise rate. We also give a matching lower bound by showing that no efficient black-box boosting algorithm can boost accuracy beyond the noise rate (assuming that one-way functions exist). Finally, we consider a variant of the standard scenario for boosting in which the “weak learner” satisfies a slightly stronger condition than the usual weak learning guarantee. We give an efficient algorithm in this framework which can boost to arbitrarily high accuracy in the presence of classification noise.