Online Learning: Stochastic, Constrained, and Smoothed Adversaries

Simple item page
dc.contributor.authorRakhlin, Alexander
dc.contributor.authorSridharan, Karthik
dc.contributor.authorTewari, Ambuj
dc.date2023-05-17T15:04:36.000
dc.date.accessioned2023-05-23T03:36:42Z
dc.date.available2023-05-23T03:36:42Z
dc.date.issued2011-01-01
dc.date.submitted2016-07-20T12:35:11-07:00
dc.description.abstractLearning theory has largely focused on two main learning scenarios: the classical statistical setting where instances are drawn i.i.d. from a fixed distribution, and the adversarial scenario whereby at every time step the worst instance is revealed to the player. It can be argued that in the real world neither of these assumptions is reasonable. We define the minimax value of a game where the adversary is restricted in his moves, capturing stochastic and non-stochastic assumptions on data. Building on the sequential symmetrization approach, we define a notion of distribution-dependent Rademacher complexity for the spectrum of problems ranging from i.i.d. to worst-case. The bounds let us immediately deduce variation-type bounds. We study a smoothed online learning scenario and show that exponentially small amount of noise can make function classes with infinite Littlestone dimension learnable.
dc.identifier.urihttps://repository.upenn.edu/handle/20.500.14332/47847
dc.legacy.articleid1141
dc.legacy.fulltexturlhttps://repository.upenn.edu/cgi/viewcontent.cgi?article=1141&context=statistics_papers&unstamped=1
dc.source.issue459
dc.source.journalStatistics Papers
dc.source.journaltitleAdvances in Neural Information Processing Systems
dc.source.statuspublished
dc.source.volume24
dc.subject.otherStatistics and Probability
dc.titleOnline Learning: Stochastic, Constrained, and Smoothed Adversaries
dc.typePresentation
digcom.contributor.authorRakhlin, Alexander
digcom.contributor.authorSridharan, Karthik
digcom.contributor.authorTewari, Ambuj
digcom.identifierstatistics_papers/459
digcom.identifier.contextkey8858982
digcom.identifier.submissionpathstatistics_papers/459
digcom.typeconference
dspace.entity.typePublication
upenn.schoolDepartmentCenterStatistics Papers
Files
Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
4262_online_learning_stochastic_constrained_and_smoothed_adversaries.pdf
Size:
159.2 KB
Format:
Adobe Portable Document Format
Download
Collection
Presentations