Bayesian voting schemes and large margin classifiers

We provide a novel theoretical analysis of such classi ers, based on data-dependent VC theory, proving that they can be expected to be large margin hyperplanes in a Hilbert space, and hence to have low e ective VC-dimension. This not only explains the remarkable resistance to over tting exhibited by such classi ers, but also colocates them in the same class as other systems, such as Support Vector Machines and Adaboost, which have a similar performance.