A Multiclass Extension To The Brownboost Algorithm

Brownboost is an adaptive, continuous time boosting algorithm based on the Boost-by-Majority (BBM) algorithm. Though it has been little studied at the time of writing, it is believed that it should prove especially robust with respect to noisy data sets. This would make it a very useful boosting algorithm for real-world applications. More familiar algorithms such as Adaboost, or its successor Logitboost, are known to be especially susceptible to over tting the training data examples. This can lead to a poor generalization error in the presence of class noise, since weak hypotheses induced at later iterations to t the noisy examples will tend to be given undue in uence in the nal combined hypothesis. Brownboost allows us to specify an expected base-line error rate in advance, corresponding to our prior beliefs about the proportion of noise in the training data, and thus avoid over tting. The original derivation of Brownboost is restricted to binary classi cation problems. In this paper we propose a natural multi-class extension to the basic algorithm, incorporating error-correcting output codes and a multi-class gain measure. We test two-class and multi-class versions of the algorithm on a number of real and simulated data sets with articial class noise, and show that Brownboost consistently outperforms Adaboost in these situations.