Show simple item record

dc.contributor.authorFrogner, Charlieen_US
dc.contributor.authorClaici, Sebastianen_US
dc.contributor.authorChien, Edwarden_US
dc.contributor.authorSolomon, Justinen_US
dc.date2021-01-04
dc.date.accessioned2021-11-03T14:14:48Z
dc.date.available2021-11-03T14:14:48Z
dc.identifier.citationC. Frogner, S. Claici, E. Chien, J. Solomon. "Incorporating Unlabeled Data into Distributionally-Robust Learning." Journal of Machine Learning Research,
dc.identifier.issn1532-4435
dc.identifier.urihttps://hdl.handle.net/2144/43260
dc.description.abstractWe study a robust alternative to empirical risk minimization called distributionally robust learning (DRL), in which one learns to perform against an adversary who can choose the data distribution from a specified set of distributions. We illustrate a problem with current DRL formulations, which rely on an overly broad definition of allowed distributions for the adversary, leading to learned classifiers that are unable to predict with any confidence. We propose a solution that incorporates unlabeled data into the DRL problem to further constrain the adversary. We show that this new formulation is tractable for stochastic gradient-based optimization and yields a computable guarantee on the future performance of the learned classifier, analogous to—but tighter than—guarantees from conventional DRL. We examine the performance of this new formulation on 14 real data sets and find that it often yields effective classifiers with nontrivial performance guarantees in situations where conventional DRL produces neither. Inspired by these results, we extend our DRL formulation to active learning with a novel, distributionally-robust version of the standard model-change heuristic. Our active learning algorithm often achieves superior learning performance to the original heuristic on real data sets.en_US
dc.language.isoen_US
dc.publisherMicrotome Publishingen_US
dc.relation.ispartofJournal of Machine Learning Research
dc.subjectDistributionally robust optimizationen_US
dc.subjectWasserstein distanceen_US
dc.subjectOptimal transporten_US
dc.subjectSupervised learningen_US
dc.subjectActive learningen_US
dc.titleIncorporating unlabeled data into distributionally-robust learningen_US
dc.typeArticleen_US
dc.description.versionAccepted manuscripten_US
pubs.elements-sourcemanual-entryen_US
pubs.organisational-groupBoston Universityen_US
pubs.organisational-groupBoston University, College of Arts & Sciencesen_US
pubs.organisational-groupBoston University, College of Arts & Sciences, Department of Computer Scienceen_US
pubs.publication-statusAccepteden_US
dc.identifier.mycv617043


This item appears in the following Collection(s)

Show simple item record