I kinda tune out when papers start talking about hessians in deep learning.