| | |
Summary: Empirical Bernstein Stopping
Volodymyr Mnih mnih@cs.ualberta.ca
Csaba Szepesv´ari szepesva@cs.ualberta.ca
Department of Computing Science, University of Alberta, Edmonton, AB T6G 2E8 Canada
Jean-Yves Audibert audibert@certis.enpc.fr
Certis - Ecole des Ponts, 6 avenue Blaise Pascal, Cit´e Descartes, 77455 Marne-la-Vall´ee France
Willow - ENS / INRIA, 45 rue d'Ulm, 75005 Paris, France
Abstract
Sampling is a popular way of scaling up ma-
chine learning algorithms to large datasets.
The question often is how many samples
are needed. Adaptive stopping algorithms
monitor the performance in an online fash-
ion and they can stop early, saving valu-
able resources. We consider problems where
probabilistic guarantees are desired and
demonstrate how recently-introduced empir-
ical Bernstein bounds can be used to design
stopping rules that are efficient. We provide
upper bounds on the sample complexity of
|
