On the stability of bootstrap estimators for support vector
machines
ABSTRACT
Let T be an estimator based on a continuous operator from the space of
probability measures over a compact metric space into a complete
separable metric space. In this talk we show that bootstrap
approximations to the distribution of T are stable ("qualitatively
robust"). An application of this result that is of particular interest
is the bootstrap approximation to the distribution of support vector
machines. This is a joint work with Andreas Christmann and
Stefan van Aelst.