The professor let this go for a moment, then began to rap his knuckles on the podium until the room quieted again.

On the acoustic side, we use a score fusion of three strong models: recurrent nets with maxout activations, very deep convolutional nets with 3x3 kernels, and bidirectional long-short term memory nets which operate on bottleneck features.

Surely this was the face of a desperado.

I can't make it to the party, but I'll be there in spirit.