深度学习参考文献

Improvinginformationflowindeepnetworkshelpstoeasethetrainingdifficultiesandutilizeparametersmoreefficiently.Hereweproposeanewconvolutionalneuralnetworkarchitecturewithalternatelyupdatedclique(CliqueNet).Incontrasttopriornetworks,therearebothforwardandbackwardconnectionsbetweenanytwolayer