NLP and Deep Learning: introduction - Session I

In this session, Roelof Pieters will present about NLP and Deep Learning. This will be the first session of series of talks about Deep Learning. Roelof is a PhD student at the CSC department at KTH. He is also CTO of a small startup, Feeda. His research focuses on distributed language models and graph-based approaches for NLP and Deep Learning.

"Deep Learning" has become a "buzz" word recently. In this first of a series of upcoming sessions - if the interest is there - we will start taming the beast !
The session will consist of two parts:

Part #1:
In part 1 we will give some first definitions, and how deep approaches contrast with more classical machine learning approaches. We will delve (very shortly) into the history of Deep Learning, the fields it is currently being applied to, and the current state, as well as future possible directions of Deep Learning, specifically for NLP.

• coffee/beer break -

Part #2:
In part 2 we will put some meat on the theoretical bones, and with the help of some actual code examples (everyone likes code!) we will go over the main deep architectures, like Deep belief Networks (DBN), Convolutional Nets (CNN), and - especially relevant for NLP - Recurrent Neural Networks (RNN). If time allows, Roelof will go through some examples and demos.

--------------

Schedule

--------------

18:15-1840 Mingling and intro.

18:40-18:45 Welcome

18:45-19:25 Part one of talk: intro + theory

19:25-19:45 Coffee break

19:45-20:30 Part 2 of talk: some DL architectures, example code, interactive demos + questions/discussion

20:30-21:00 Mingling