SampleAmplification:IncreasingDatasetSizeEvenwhenLearningisImpossibleBrianAxelrod1ShivamGarg1VatsalSharan1GregoryValiant1Abstractordoesitsufficetohaveaccesstoasmallerdatasetofsizen<mdrawnfromD,andt...
BeingBayesian,EvenJustaBit,FixesOverconfidenceinReLUNetworksAgustinusKristiadi1MatthiasHein1PhilippHennig12Abstract2016).Themostobviouscasesaretestpointsthatlie“faraway”fromthetrainingdata.Manyme...