As the new tutorial chair for the next Euroscipy conference, I was thinking this January about the difficult choice of possible topics for advanced tutorials. Euroscipy is the annual European conference on Scientific Python; it is organized with two days of tutorials and two days of conference (keynotes and contributed talks), with attendees from academia, industry and start-ups.

While tutorials of the introduction track always covers a core of packages and skills that are essential to beginners, advanced users have enjoyed quite a large variety of topics during the advanced tutorial track of the last conference editions. Attendees have various backgrounds and expectations, from getting bleeding-edge news about the latest cool packages to improving their numerical skills, learning about coding best practices, or simply using core packages in a more efficient and advanced way. How to find the right balance between topics answering these different needs? A friend then suggested: hey, why not ask the attendees themselves what they would like to learn during tutorials?

I quickly set up a list of potential topics, wrote a simple form, and called for opinions on Twitter two months ago. I was lucky that this message was retweeted enough, and that people seem to care about this topic: the form got completed 71 times, which represents about half a typical audience for the advanced tutorials of Euroscipy. Quite a good score therefore!

The results of the poll (which is now closed) can be seen here. Note that some topics were not included in the list, since they had been presented the year before (like my favorite subject scikit-image).

Since several outcomes of the poll were quite counter-intuitive to me, I'm really glad I got so much feedback from an audience that I hope is quite diverse -- of course there must be some bias due to the use of Twitter for communication, but I still hope that the poll is a good proxy for what people would like to see.

A few lessons from this poll

• advanced tutorials on core packages or skills, such as advanced NumPy, Python, and scientific plotting, all score very high.

• software development methods (how to turn your bunch of scripts into an efficient and reliable package that you can share with others) also look quite popular, as shown by the high scores of Testing, Packaging and Parallel computing.

• machine-learning-related topics, such as scikit-learn or neural nets/deep learning, were popular as well.

• version control did not seem a priority to poll respondents -- let's hope it means that most scientists now master version control (although this would somewhat contradict my experience!).

• unsurprisingly, topics that are slightly more specialized, such as natural language processing or linear algebra, were not as popular as less specialized subjects. I might pick one or two of these topics that are of high interest to a fraction of the audience, to beat the tyranny of the majority (of course, a poll is meant to obtain what the majority thinks!).

So, what's next? Now I will be looking for speakers for the most popular topics revealed by the poll, plus a couple of others for which I'll claim a wild card privilege as tutorial chair :-). After looking at the poll results, if you feel like giving a tutorial at Euroscipy 2016 (Erlangen, Germany, August 23-27), please contact me, preferably by e-mail (firtsname.lastname@nsup.org). There should also be a more official call for talks and tutorials on the Euroscipy website in a couple of days/weeks.

If you feel like commenting on the results of the opinion poll, please leave a comment on this blog post. I would like to finish with a big thank you to all the people who took the time to complete the opinion poll: it really helped!