Distilling the Knowledge in a Neural Network

   Abstract