PolicyInformationCapacity:Information-theoreticMeasureforTaskComplexityinDeepReinforcementLearningHirokiFuruta1TatsuyaMatsushima1TadashiKozuno2YutakaMatsuo1SergeyLevine3OfirNachum3ShixiangShaneGu3A...
Information-theoreticLocalMinimaCharacterizationandRegularizationZhiweiJia1HaoSu1Abstractinitionsof“flatness/sharpness”havebeenintroducedandanalyzed(Keskaretal.,2017;Neyshaburetal.,2018;2017;Rece...
TheInformation-theoreticValueofUnlabeledDatainSemi-SupervisedLearningAlexanderGolovnev1Da´vidPa´l2Bala´zsSzo¨re´nyi2Abstractofalgorithmsindexedbythe(uncountablymany)distri-butionsoverthedomain...
Information-theoreticConsiderationsinBatchReinforcementLearningJinglinChen1NanJiang1AbstractwhentheyworkiscentraltoourunderstandingofRL.Ex-istingworksthatanalyzeerrorpropagationandfinitesam-Value-f...
LearningtoExplain:AnInformation-theoreticPerspectiveonModelInterpretationJianboChen12LeSong34MartinJ.Wainwright15MichaelI.Jordan1Abstractisfeatureselection,whichselectsasubsetoffeaturesthatareusefu...
FastInformation-theoreticBayesianOptimisationBinxinRu1MarkMcLeod1DiegoGranziol1MichaelA.Osborne12Abstractofproblemsaregenerallynon-convexandtheirclosed-formexpressionsandderivativesareunknown(Shahr...