AModularAnalysisofProvableAccelerationviaPolyak’sMomentum:TrainingaWideReLUNetworkandaDeepLinearNetworkJun-KunWang1Chi-HengLin2JacobAbernethy1Abstract1.IntroductionIncorporatingaso-called“momentu...
RecursiveSketchesforModularDeeplikeouralgorithmstobesmall,efficient,androbust.ThisLearningworkaimstofurtherexploreonesuchsub-question:canwedesignasystemontopofneuralnetsthatefficientlystoresBadihGh...
CURIOUS:IntrinsicallyMotivatedModularMulti-GoalReinforcementLearningCe´dricColas1PierreFournier2OlivierSigaud2MohamedChetouani2Pierre-YvesOudeyer1AbstractFigure1.TheModularGoalFetchArmenvironment....
ModularMultitaskReinforcementLearningwithPolicySketchesJacobAndreas1DanKlein1SergeyLevine1Abstractτ1:makeplanksΠ1τ2:makesticksΠ2b1:getwoodK1π1Wedescribeaframeworkformultitaskdeepre-b2:useworkb...