Please use this identifier to cite or link to this item:
DC FieldValue
dc.titleActive learning for probabilistic hypotheses using the maximum Gibbs error criterion
dc.contributor.authorNguyen, V.C.
dc.contributor.authorLee, W.S.
dc.contributor.authorYe, N.
dc.contributor.authorChai, K.M.A.
dc.contributor.authorChieu, H.L.
dc.identifier.citationNguyen, V.C.,Lee, W.S.,Ye, N.,Chai, K.M.A.,Chieu, H.L. (2013). Active learning for probabilistic hypotheses using the maximum Gibbs error criterion. Advances in Neural Information Processing Systems. ScholarBank@NUS Repository.
dc.description.abstractWe introduce a new objective function for pool-based Bayesian active learning with probabilistic hypotheses. This objective function, called the policy Gibbs error, is the expected error rate of a random classifier drawn from the prior distribution on the examples adaptively selected by the active learning policy. Exact maximization of the policy Gibbs error is hard, so we propose a greedy strategy that maximizes the Gibbs error at each iteration, where the Gibbs error on an instance is the expected error of a random classifier selected from the posterior label distribution on that instance. We apply this maximum Gibbs error criterion to three active learning scenarios: non-adaptive, adaptive, and batch active learning. In each scenario, we prove that the criterion achieves near-maximal policy Gibbs error when constrained to a fixed budget. For practical implementations, we provide approximations to the maximum Gibbs error criterion for Bayesian conditional random fields and transductive Naive Bayes. Our experimental results on a named entity recognition task and a text classification task show that the maximum Gibbs error criterion is an effective active learning criterion for noisy models.
dc.typeConference Paper
dc.contributor.departmentCOMPUTER SCIENCE
dc.description.sourcetitleAdvances in Neural Information Processing Systems
Appears in Collections:Staff Publications

Show simple item record
Files in This Item:
There are no files associated with this item.

Page view(s)

checked on Mar 29, 2020

Google ScholarTM


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.