GradientDescentLearnsOne-hidden-layerCNN:Don’tbeAfraidofSpuriousLocalMinimaSimonS.Du1JasonD.Lee2YuandongTian3Barnaba´sPo´czos1AartiSingh1Abstractfully.WhysuchsimplemethodsinlearningDCNNissuc-ces...
RecoveryGuaranteesforOne-hidden-layerNeuralNetworks∗KaiZhong1ZhaoSong2PrateekJain3PeterL.Bartlett4InderjitS.Dhillon5Abstract1.IntroductionInthispaper,weconsiderregressionproblemsNeuralNetworks(NNs...