ExpandNets: Linear Over-parameterization to Train Compact Convolutional Networks

Open in new window