Spring school "Structural Inference" 2015. Overview

posted by Nikolay Baldin on April 8, 2015

Inspired by the time spent at the spring school in statistics on the German island Sylt, I decided to highlight some features of the school and collect suggestions. Below there are general remarks, kudos, my false expectation and pics.

Looking forward to the next year's school.

General comments

The school was organized by Ulrike von Luxburg and Vladimir Spokoiny and was sponsored by Research Unit 1735 . There were three short mini courses given on statistical learning by Sasha Rakhlin from UPenn, nonparametric Bayes by Judith Rousseau from Paris Dauphine and high-dimensional inference by Martin Wainwright from Berkeley. The slides and exercise sheets are available here. Personally, it was my third school organized by those people, so I know what to compare it with :).

As in the previous years, most of the participants came from Germany, France, Italy, Russia, UK. The background of people had a large variance: there were some already finishing PhD students with solid publications and, at the same time, master's students who basically just started research (I couldn't find a logical tendency in people the organizers invited).

Apart from the lectures, a social activity was organized: there was a soccer game (many kudos for that!) and an excursion to a nearby area.



  1. Exercise sessions. It was really beneficial for understanding the lectures. Many kudos to the lecturers for creating good exercise sheets.
  2. Program. It was quite diverse and at the same time really well-interconnected.
  3. Students presentations. I find it absolutely the must asking all participants (including seniors) to introduce themselves in 2-3 minutes. At this time, it was organized perfectly.
  4. Location. I really appreciate this kind of secluded places where one can focus on intensive learning with no distractions.

Food for thought

  1. Exercises. As I already mentioned, the idea with exercises was extremely nice, but, to my big surprise, not everyone was active in doing them :D. So it happened many times that a lecturer was compelled to solve them himself. Next time, we should definitely encourage people to participate more. For example, we could divide participants randomly into groups and organize a competition among groups.
  2. I believe there could be more scientific activity organized. The main objective here is to engage people in intensive learning of the material. I don't have an exact understanding of this point yet. Just have a feeling that the existing format of lectures/posters can be improved :)
  3. Good Internet connection. At least, at the seminars rooms.. Please.. All the features* with the Internet should be mentioned in advance before the school starts.
  4. Technical questions. At least, one printer should be always available.
  5. tbw

Of course, few pics

Trail to the beach

Participants shacks

View on the campus from the beach

Russian team

Sand dunes nearby