Logo Utrecht University

Methodology & Statistics Data Science Lab


02/11/2017 – A Primer on Deep Learning

We had a full house this time! Thanks all for joining. As discussed near the end of the meeting, head over to keras.rstudio.com and try out the Reuters example (almost at the bottom). In the upcoming meeting about software, we will discuss this further.

The presentation can be downloaded from here (.html download, 19MB)




What are neural networks? What is deep learning? What is everybody getting so excited about? What are some common deep learning “architectures” (models) and what are they used for? Why is deep learning hard?  And should I use deep learning?

In approximately two slides per question, our own Daniel Oberski is going to discuss these issues at a conceptual level. We’ll also do a small exercise that demonstrates the inner workings of a “deep feedforward” architecture: bring your laptop with R and Rstudio. It also helps to have the keras package installed and working (https://keras.rstudio.com/ – the second part can be harder than the first).

You must be logged in to post a comment.