TowardsUnderstandingtheDynamicsoftheFirst-OrderAdversariesZhunDeng1HangfengHe2JiaoyangHuang3WeijieJ.Su2Abstractattackssuchasinputde-noising(Guoetal.,2017),random-izedsmoothing(Ilyasetal.,2019),grad...
TowardsAdaptiveResidualNetworkTraining:ANeural-ODEPerspectiveChengyuDong1LiyuanLiu2ZichaoLi1JingboShang1Abstractisnotnew(Changetal.,2017;Wenetal.,2019),butthedynamicsofgrowingispoorlyunderstood.Inp...
Towardsnon-parametricdriftdetectionviaDynamicAdaptingWindowIndependenceDriftDetection(DAWIDD)FabianHinder1Andre´Artelt1BarbaraHammer1Abstractsuchsettings,life-longmodeladaptationratherthanclas-sic...
TowardsAccuratePost-trainingNetworkQuantizationviaBit-SplitandStitchingPeisongWang1QiangChen1XiangyuHe1JianCheng1AbstracthardwarelikeTPUorgeneralhardwarelikeCPUandGPU.Byturningthefloating-pointvalu...
TowardsaGeneralTheoryofInfinite-WidthLimitsofNeuralClassifiersEugeneA.Golikov1Abstractofinfinitewidth.Inparticular,(Jacotetal.,2018)showedthatifweightsareparameterizedinacertainwaythentheObtainingt...
MoNet3D:TowardsAccurateMonocular3DObjectLocalizationinRealTimeXichuanZhou1YicongPeng1ChunqiaoLong1FengboRen2CongShi1AbstractRedmonetal.,2016;Liuetal.,2016).However,despiteitsgreatadvantagesintermso...
DynamicKnapsackOptimizationTowardsEfficientMulti-ChannelSequentialAdvertisingXiaotianHao1ZhaoqingPeng2YiMa1GuanWang3JunqiJin2JianyeHao1ShanChen2RongquanBai2MingzhouXie2MiaoXu2ZhenzheZheng4ChuanYu2H...
BreakingtheCurseofSpaceExplosion:TowardsEffcientNASwithCurriculumSearchYongGuo1YaofoChen12YinZheng3PeilinZhao4JianChen15JunzhouHuang4MingkuiTan1AbstractStandardNASwithNASwithcurriculumsearchfixedse...
AR-DAE:TowardsUnbiasedNeuralEntropyGradientEstimationJaeHyunLim12AaronCourville1234ChristopherPal154Chin-WeiHuang12Abstractcontrolthisquantityaspartoftheoptimizationobjective.Inlightofthis,wepropos...
TowardsUnderstandingKnowledgeDistillationMaryPhuong1ChristophH.Lampert1Abstractdistillation-basedtraininghasbeenconfirmedseveraltimes:theoptimizationstepisgenerallymorewell-behavedthanKnowledgedist...
TowardsAccurateModelSelectioninDeepUnsupervisedDomainAdaptationKaichaoYou12XimeiWang12MingshengLong12MichaelI.Jordan3Abstractnaturalinrecognitiontaskswheredeepmodelshaveshowntheirsuperiority(Longet...
TowardsaUnifiedAnalysisofRandomFourierFeaturesZhuLi1Jean-FrançoisTon1DinoOglic2DinoSejdinovic1Abstractimplicitcomputationofaninnerproductbetweenrichfea-turerepresentationsofdatathroughthekerneleva...
TowardsaDeepandUnifiedUnderstandingofDeepNeuralModelsinNLP1ChaoyuGuan2XitingWang2QuanshiZhang1RunjinChen1DiHe3XingXie2Abstract(a)Gradient-basedmethod(b)OursWedefineaunifiedinformation-basedmeasureF...
PolicyCertificates:TowardsAccountableReinforcementLearningChristophDann1LihongLi2WeiWei2EmmaBrunskill3Abstractploration.Evensharpdropsinpolicyperformanceduringlearningarecommon,e.g.,whentheagentsta...
NAS-Bench-101:TowardsReproducibleNeuralArchitectureSearchChrisYing1AaronKlein2EstebanReal1EricChristiansen1KevinMurphy1FrankHutter2Abstractturesasanoptimizationproblem(Bakeretal.,2017;Zoph&Le,2016;...
ME-Net:TowardsEffectiveAdversarialRobustnesswithMatrixEstimationYuzheYang1GuoZhang1DinaKatabi1ZhiXu1AbstractForinstance,thestate-of-the-artdefensecanonlyachievelessthan50%adversarialaccuracyfor∞pe...
DBSCAN++:TowardsfastandscalabledensityclusteringJenniferJang1HeinrichJiang2AbstractSCANcanruninO(nlogn)whenthedimensionisatmost2,itquicklystartstoexhibitquadraticbehaviorinhighdi-DBSCANisaclassical...
AnalogiesExplained:TowardsUnderstandingWordEmbeddingsCarlAllen1TimothyHospedales1Abstractofwordwi.Thisenablesanalogicalquestionssuchas“manistokingaswomanisto..?”tobesolvedbyvectoraddi-Wordembeddi...
TowardsMoreEfficientStochasticDecentralizedLearning:FasterConvergenceandSparseCommunicationZebangShen12AryanMokhtari3TengfeiZhou1PeilinZhao4HuiQian1Abstractvariableofnoden,theproblemofinterestisRec...
TowardsFastComputationofCertifiedRobustnessforReLUNetworksTsui-WeiWeng1HuanZhang2HonggeChen1ZhaoSong34Cho-JuiHsieh2DuaneBoning1InderjitS.Dhillon4LucaDaniel1Abstract1.IntroductionVerifyingtherobustn...