AcceleratingLarge-ScaleInferencewithAnisotropicVectorQuantizationRuiqiGuo1PhilipSun1ErikLindgren1QuanGeng1DavidSimcha1FelixChern1SanjivKumar1Abstract(MIPS)problem,consideradatabaseX={xi}i=1,2,...,n...
TheAnisotropicNoiseinStochasticGradientDescent:ItsBehaviorofEscapingfromSharpMinimaandRegularizationEffectsZhanxingZhu123JingfengWu1BingYu1LeiWu1JinwenMa1Abstract90Understandingthebehaviorofstochas...