Man, workshops that start at 7.30am are not much fun… At least, when one isn’t getting much sleep beforehand: stage fright! After far too little sleep, I (sleep)walked to the venue of the workshop and realized that all of the people had some nice badges. Duh! Had no idea where to register and my room didn’t pass by the lobby, so I completely forgot about it. Well, it was at least still enough time to do that before the workshop officially started with an introduction by Soeren.
The first invited speaker was John Eaton, the lead developer of Octave. It is sad to hear that even such an established project like Octave cannot obtain funding in order to employ a single person (John is no longer working for UW Madison). If you ever want to see a software that got rewritten several times, in the search for the highest optimization, then take a look at Torch, written in Lua and C/C++ (all versions are apparently incompatible with each other). After a few other talks, the discussion about a data interchange format between all those different approaches and frameworks was, let’s say, rather interesting. There were at least as many opinions in the room as there were people. Maybe even more! Interestingly enough, Mikio introduced Weka‘s ARFF format as an initial rudimentary starting point (with readers and writers in different programming languages). Man, he almost got lynched! 😉 Other people suggested HDF5, CDF or EBML. All of the latter ones seem a bit like an overkill to me…
A lift just around the corner from my hotel. 🙂
After the lunch break (quite a few of us went to lunch together), John Hunter was kicking off the afternoon session as invited speaker. He is the lead developer of matplotlib, a very cool 2-D plotting library for Python. Was quite surprised what you could do with this library (just check out the gallery!). The talk directly after that, about Disco, was very interesting. After a few more talks, it was finally my turn to give my two 5 minute presentations about Experiment Databases for Machine Learning (actually Joaquin‘s talk) and BenchMarking Via Weka. Man, was I nervous… And drinking all that water during those other talks didn’t make things easier, if you know what I mean. 😉 But, it wasn’t a complete disaster, so I’m quite happy! I wonder, when the talks will be up on videolectures.net. The discussion about reproducibility, kicked off by Cheng this time, was as lively as the previous one… But it’s nice to see people getting involved. And once again, people couldn’t agree on anything.
Of course, we had to go out for dinner to a steak house!! I went for the caesar salad, being the only vegetarian at the table. Was a really cool night after a successful workshop day. 🙂