I knew next to nothing about the country of Azerbaijan or its capital city, Baku, before my visit. My guidebook describes Baku as “the architectural love child of Paris and Dubai, with plenty of Soviet genes floating half-hidden in the background. Nowhere else do East and West blend as seamlessly and as chaotically.” That’s a bit dramatic, but perhaps not too far off the mark: In my brief stay I found the city and fascinating and unique.
My host at ADA University takes the prize for most enthusiastic to date. The run-up to my visit included over-the-top advertising and numerous exuberant emails. Once I arrived I found the participants to be as enthusiastic as the host, shouting out their guesses at answers even before I asked the questions, lining up to chat with me at every break, and taking more photos than their counterparts in Southeast Asia, which I hadn’t thought possible. (Incidentally, it’s a matter of debate whether Azerbaijan is part of Asia or Europe.) I’ve been keeping up a very modest social media stream along with this blog: an Instagram feed with one captioned photo per day, and a Facebook page that simply reposts the daily photos along with links to this blog. In Azerbaijan not only were the host and participants enthusiastic in person, but my posts experienced a social media explosion. I should be clear that “explosion” is relative to the starting point: I went from 10–20 “likes” on each Instagram photo to a whopping 50+, and from 200–300 “people reached” on Facebook posts to over 2000. It’s still a long way to social media stardom.
In addition to the by now standard Big Data short-course, the very active Azerbaijan chapter of the ACM professional society for computing scheduled its annual half-day conference for women in technology to coincide with my visit. I joined a panel of Azerbaijani women who mostly began as mathematicians and then switched to various types of tech. The discussion was lively, with many questions from the audience and stories from the panelists. As the panel came to a close, the women very much wanted to make sure I didn’t leave with a poor impression of Azerbaijani men. As one woman said: “It’s Saturday, we’re all here and our husbands are at home with the kids. Who can complain about that?” The proceedings were conducted in Azeri (a standard nickname for the Azerbaijani language), so I was provided with a professional “whispered interpreter.” He was remarkably effective at simultaneous translation — Azeri in and English out, fully parallel without missing a beat — allowing me to feel completely immersed in the discussion.
My visit had just one annoyance, and an unexpected one: problems with basic technology. I’ve had so little trouble connecting my laptop to whatever form of projection I find at one remote destination after another that I’ve pretty much stopped worrying about it. (I do carry an arsenal of cables, connectors, adapters, etc., but I’ve always been able to produce the right combination without too much fuss.) ADA University was the first place with serious difficulties, and it wasn’t because of too little technology, it was too much! I’m absolutely convinced the world is better off with trusty old projectors and pull-down screens than with “multi-mode tiled smart-panels.” We got past each of the several hurdles eventually, but it cost us an hour or two of teaching time.
I’ve been asked several times what datasets I use for hands-on teaching of Big Data. The truth is that only Small Data is manageable with this number of students in such a wide variety of settings, but I’m very careful to ensure that all of the tools and techniques we use apply equally well to much bigger datasets. (For example, when we work with spreadsheets — which, incidentally, are surprisingly powerful for data analysis — students aren’t allowed to scroll through the data, since that approach wouldn’t scale to millions of data items.) Primarily we use two datasets: European city temperatures and soccer world cup statistics. Over the short-course we:
- Learn many different ways of discovering and visualizing the inverse correlation between latitude and temperature (when latitude increases temperature decreases, generally speaking).
- Find out that, in the 2010 world cup, defenders made twice as many passes on average as forwards, that the overall winner (Spain) also had the overall highest number of passes per minute, and that Lionel Messi’s performance was an outlier.
- Calculate that cities in the European Union are warmer on average than non-EU cities by a small margin, but the gap will increase after Brexit.
- Discover that “decision trees” do better than “random forests” at predicting a soccer player’s position from his performance (measured in minutes, passes, tackles, shots, and saves), and that “naive Bayes” predicts city temperatures best using latitude only, while other methods do better using both longitude and latitude. (Machine learning folks: take note!)
Although I’ve been using the same datasets for several months now, the Azerbaijanis were the first to point out an error: The country of Bosnia and Herzegovina is listed as having no coastline, but if you look very closely it boasts a tiny bit.
On to Mauritius, my last stop for now. I’ll miss the instructional odyssey. It was great to stop through a place like Azerbaijan — one I knew nothing about — and spend a few days with folks who were immensely appreciative of my visit and teaching.