On the Efficient Representation and Execution of Deep Acoustic Models

   Abstract