We derive two oracle inequalities for regularized boosting algorithms for classification. The first oracle inequality generalizes and refines a result from [4], while the second oracle inequality leads to faster learning rates than those of [4] whenever the set of weak learners does not perfectly approximate the target function. The techniques leading to the second oracle inequality are based on the wellknown approach of adding some artificial noise to the labeling process.
I. Steinwart, Two Oracle Inequalities for Regularized Boosting Classifiers. Statistics and Its Interface, to appear. Los Alamos National Laboratory Technical Report LA-UR-08-7206. [ Abstract | PDF (297 KB) ]
