arXiv Analytics

Sign in

arXiv:2001.06105 [cs.LG]AbstractReferencesReviewsResources

Better Boosting with Bandits for Online Learning

Nikolaos Nikolaou, Joseph Mellor, Nikunj C. Oza, Gavin Brown

Published 2020-01-16Version 1

Probability estimates generated by boosting ensembles are poorly calibrated because of the margin maximization nature of the algorithm. The outputs of the ensemble need to be properly calibrated before they can be used as probability estimates. In this work, we demonstrate that online boosting is also prone to producing distorted probability estimates. In batch learning, calibration is achieved by reserving part of the training data for training the calibrator function. In the online setting, a decision needs to be made on each round: shall the new example(s) be used to update the parameters of the ensemble or those of the calibrator. We proceed to resolve this decision with the aid of bandit optimization algorithms. We demonstrate superior performance to uncalibrated and naively-calibrated on-line boosting ensembles in terms of probability estimation. Our proposed mechanism can be easily adapted to other tasks(e.g. cost-sensitive classification) and is robust to the choice of hyperparameters of both the calibrator and the ensemble.

Related articles: Most relevant | Search more
arXiv:1711.03343 [cs.LG] (Published 2017-11-09)
Analysis of Dropout in Online Learning
arXiv:2009.11942 [cs.LG] (Published 2020-09-24)
Online Learning With Adaptive Rebalancing in Nonstationary Environments
arXiv:2007.05665 [cs.LG] (Published 2020-07-11)
A Computational Separation between Private Learning and Online Learning