1:15pm - 2:30pm in Wilson 117
Towards Understanding Long Short Term Memory Networks
We discuss early-stage research towards understanding LSTMs through interpretations of state-space models, in particular hidden Markov models. Ultimately, work in this direction aims to enhance our understanding of deep learning systems and aid our ability to map the state of the system onto domain knowledge. The talk will be self-contained, and only assumes a basic familiarity with statistics and probability.
Date:
Monday, January 28, 2019