It was nice seeing a bunch of our group members at the applied ml day. This is a good opportunity to cover two motifs that came up in the talks: (I) Features (features, features..); (II) Computational power.
Below is a list of optional papers for our session. We'll pick only one/two of them. Please join the effort (by covering even a modest portion, e.g., a section).
Learning features - This deep learning tutorial covers the essentials of this field. It's about time we'll touch this fascinating field. http://bit.ly/19HPkOC
(Recycling a few works we wanted to cover in previous sessions, yet, didn't get to.)
Aggressive feature selection - The why, when, & how of getting rid of many (redundant) features. I think that Kira mentioned such situation in her talk. http://bit.ly/1aE2SJd
How to do massive computations on a small scale machine - Danny mentioned in his talk the power of being 'computationaly-wise.' The following 'graph-chi' paper is all about engineering. I think it is a gem. http://bit.ly/1mkCfuS
Learning behavioural dynamics on the web: We haven't had the time to cover this paper last time (in our time-series session). Let's look at it from a "feature engineering" point of view. http://bit.ly/18uiPTx