A Gift from Knowledge Distillation: Fast Optimization, Network Minimization and Transfer Learning | Synapse