Gradient Descent Finds Global Minima of Deep Neural Networks

GradientDescentFindsGlobalMinimaofDeepNeuralNetworksSimonS.Du1JasonD.Lee2HaochuanLi34LiweiWang54XiyuZhai6AbstractThesecondmysteriousphenomenonintrainingdeepneuralnetworksis“deepernetworksareharder...