UnderstandingSelf-supervisedLearningDynamicswithoutContrastivePairsYuandongTian1XinleiChen1SuryaGanguli12Abstractquiringexpensivetargetlabels(Devlinetal.,2018).Manystate-of-the-artSSLmethodsincompu...
Self-supervisedGraph-levelRepresentationLearningwithLocalandGlobalStructureMinghaoXu1HangWang1BingbingNi1HongyuGuo2JianTang345Abstract2020).Thesemethodsareusuallytrainedinasupervisedfashion,whichre...
Self-supervisedandSupervisedJointTrainingforResource-richMachineTranslationYongCheng1WeiWang†LuJiang12WolfgangMacherey1Abstractsupervisedtaskonabundantunlabeleddata(i.e.monolin-gualsentences).Inth...
WhiteningforSelf-supervisedRepresentationLearningAleksandrErmolov1AliaksandrSiarohin1EnverSangineto1NicuSebe1Abstractofwordsinasentenceisusedtolearnalanguagemodel(Mikolovetal.,2013a;b;Devlinetal.,2...
CausalCuriosity:RLAgentsDiscoveringSelf-supervisedExperimentsforCausalRepresentationLearningSumedhASontakke1ArashMehrjou2LaurentItti1BernhardSchölkopf2Abstractform.Thus,therehasbeenrecentinteresti...
BarlowTwins:Self-supervisedLearningviaRedundancyReductionJureZbontar1LiJing1IshanMisra1YannLeCun12Ste´phaneDeny1AbstractSelf-supervisedlearning(SSL)israpidlyclosingFigure1.BARLOWTWINS’sobjectivef...
Skew-Fit:State-CoveringSelf-supervisedReinforcementLearningVitchyrH.Pong1MurtazaDalal1StevenLin1AshvinNair1ShikharBahl1SergeyLevine1AbstractFigure1.Left:RobotlearningtoopenadoorwithSkew-Fit,without...
Self-supervisedLabelAugmentationviaInputTransformationsHankookLee1SungJuHwang234JinwooShin21Abstractsentationlearningforimages(Doerschetal.,2015;Noroozi&Favaro,2016;Larssonetal.,2017;Gidarisetal.,2...
Graph-based,Self-supervisedProgramRepairfromDiagnosticFeedbackMichihiroYasunaga1PercyLiang1LSTMLASTbMstractLSTMBrokenProgramEvaluator(compiler)WeconsidertheprLoSbTlMemoflearnLiSnTgMtorepairpro-(`ch...
AutomaticShortcutRemovalforSelf-supervisedRepresentationLearningMatthiasMinderer12OlivierBachem1NeilHoulsby1MichaelTschannen1AbstractInputimageLensoutputDifferenceInSelf-supervisedvisualrepresentat...
Self-supervisedExplorationviaDisagreementDeepakPathak1DhirajGandhi2AbhinavGupta23Abstracttotheagentaresparse.Thecommonapproachtoexplo-rationhasbeentogenerate“intrinsic”rewards,i.e.,rewardsEfficie...
Curiosity-drivenExplorationbySelf-supervisedPredictionDeepakPathak1PulkitAgrawal1AlexeiA.Efros1TrevorDarrell1Abstract(a)learntoexploreinLevel-1(b)explorefasterinLevel-2Inmanyreal-worldscenarios,rew...