Fetching the paper…

ExpandNets: Linear Over-parameterization to Train Compact Convolutional Networks · Around