ECAI-2002 Conference Paper

[full paper]

A Simple Algorithm for Learning Stable Machines

Savina Andonova, André Elisseeff, Theodoros Evgeniou, Massimiliano Pontil

We present a simple algorithm for learning stable machines which is motivated by recent results in statistical learning theory. The algorithm is similar to Breiman's bagging despite some important differences in that it computes an ensemble combination of machines trained on small random subsamples of an initial training set. A remarkable property is that it is often possible to just use the empirical error of these combinations of machines for model selection. We report experiments using support vector machines and neural networks validating the theory.

Keywords: Machine Learning, Statistical Learning Theory, Bagging, Support Vector Machines, Neural Networks

Citation: Savina Andonova, André Elisseeff, Theodoros Evgeniou, Massimiliano Pontil: A Simple Algorithm for Learning Stable Machines. In F. van Harmelen (ed.): ECAI2002, Proceedings of the 15th European Conference on Artificial Intelligence, IOS Press, Amsterdam, 2002, pp.513-517.