MakingConvolutionalNetworksShift-InvariantAgainRichardZhang1Abstractpiricalevidencesuggestsmax-poolingprovidesstrongertaskperformance(Schereretal.,2010),leadingtoitsModernconvolutionalnetworksareno...
MaketheMinorityGreatAgain:First-OrderRegretBoundforContextualBanditsZeyuanAllen-Zhu1Se´bastienBubeck1YuanzhiLi2Abstract•Theadversaryselectsalossfunctiont:[K]→[0,1].Regretboundsinonlinelearningco...